[{"data":1,"prerenderedAt":5634},["ShallowReactive",2],{"guide-en-claude-code\u002Fclaude-api-pricing":3,"guide-siblings-en-claude-code":660,"guide-alt-en-claude-code\u002Fclaude-api-pricing":5631},{"id":4,"title":5,"author":6,"body":7,"date":630,"description":631,"extension":632,"faq":633,"meta":649,"navigation":650,"order":651,"path":652,"readTime":653,"seo":654,"stem":655,"topic":656,"translationId":657,"updated":658,"__hash__":659},"guides\u002Fguides\u002Fclaude-code\u002Fclaude-api-pricing.md","Claude API pricing per million tokens (2026), every model","Walma Engineering",{"type":8,"value":9,"toc":612},"minimark",[10,14,25,30,33,190,193,196,219,222,226,231,234,271,274,278,281,285,316,320,331,335,357,361,367,373,436,439,445,449,452,490,497,501,504,526,529,533,536,575,588,592],[11,12,13],"p",{},"Claude API pricing looks simple: a price per million tokens in, a price per million tokens out. What makes bills hard to predict is everything around it: which model, how much context each request carries, whether it is cached, and which tools the model calls along the way.",[11,15,16,17,24],{},"This guide lists the current prices, explains the modifiers that change them, works through two realistic examples, and ends with the levers that actually move a company's bill. Prices are from Anthropic's ",[18,19,23],"a",{"href":20,"rel":21},"https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fabout-claude\u002Fpricing",[22],"nofollow","pricing page"," as of October 2026, in US dollars. They change, so check the source before you budget.",[26,27,29],"h2",{"id":28},"claude-api-prices-per-model","Claude API prices per model",[11,31,32],{},"Prices are per million tokens (MTok). As a rough guide, a million tokens is about 750,000 English words.",[34,35,36,58],"table",{},[37,38,39],"thead",{},[40,41,42,46,49,52,55],"tr",{},[43,44,45],"th",{},"Model",[43,47,48],{},"Input",[43,50,51],{},"Output",[43,53,54],{},"Cache read",[43,56,57],{},"Typical use",[59,60,61,79,96,113,128,142,159,175],"tbody",{},[40,62,63,67,70,73,76],{},[64,65,66],"td",{},"Claude Fable 5.1",[64,68,69],{},"$10",[64,71,72],{},"$50",[64,74,75],{},"$0.25",[64,77,78],{},"The most capable model, hardest problems",[40,80,81,84,87,90,93],{},[64,82,83],{},"Claude Opus 5.5",[64,85,86],{},"$4",[64,88,89],{},"$20",[64,91,92],{},"$0.20",[64,94,95],{},"Complex reasoning, coding, agents",[40,97,98,101,104,107,110],{},[64,99,100],{},"Claude Opus 5",[64,102,103],{},"$5",[64,105,106],{},"$25",[64,108,109],{},"$0.50",[64,111,112],{},"Previous Opus generation",[40,114,115,118,121,123,125],{},[64,116,117],{},"Claude Sonnet 5.5",[64,119,120],{},"$2",[64,122,69],{},[64,124,92],{},[64,126,127],{},"The default for most production workloads",[40,129,130,133,135,137,139],{},[64,131,132],{},"Claude Sonnet 5",[64,134,120],{},[64,136,69],{},[64,138,92],{},[64,140,141],{},"Previous Sonnet generation",[40,143,144,147,150,153,156],{},[64,145,146],{},"Claude Sonnet 4.6",[64,148,149],{},"$3",[64,151,152],{},"$15",[64,154,155],{},"$0.30",[64,157,158],{},"Older Sonnet, still widely deployed",[40,160,161,164,167,169,172],{},[64,162,163],{},"Claude Haiku 5.5",[64,165,166],{},"$0.10",[64,168,109],{},[64,170,171],{},"see note",[64,173,174],{},"High-volume, simple tasks: classification, routing, extraction, subagents",[40,176,177,180,183,185,187],{},[64,178,179],{},"Claude Haiku 4.5",[64,181,182],{},"$1",[64,184,103],{},[64,186,166],{},[64,188,189],{},"Previous Haiku generation, 200K context",[11,191,192],{},"Haiku 5.5's prices apply to prompts up to 100K tokens; longer prompts cost $0.50 input \u002F $2.50 output. Check Anthropic's pricing page for its cache-read rate.",[11,194,195],{},"Three things in this table are easy to miss:",[197,198,199,207,213],"ul",{},[200,201,202,206],"li",{},[203,204,205],"strong",{},"Output costs five times input"," on every current model. A task that writes long answers costs far more than one that reads a lot and answers briefly.",[200,208,209,212],{},[203,210,211],{},"Newer is not always pricier."," Opus 5.5 is cheaper per token than Opus 5, and Sonnet 5's $2\u002F$10 introductory price became its standard price instead of rising to $3\u002F$15.",[200,214,215,218],{},[203,216,217],{},"Tokens are not words."," Claude 4.7 and later models use a newer tokenizer that produces roughly 30% more tokens for the same text. Comparing per-token prices across model generations understates that difference.",[11,220,221],{},"Claude 4.6 and later models include the full 1M-token context window at standard prices: a 900k-token request costs the same per token as a 9k-token one. It is still 900k tokens, so long context is a capability, not a discount.",[26,223,225],{"id":224},"what-changes-the-price","What changes the price",[227,228,230],"h3",{"id":229},"prompt-caching","Prompt caching",[11,232,233],{},"Most requests repeat a large prefix: the system prompt, tool definitions, a document, the conversation so far. Prompt caching stores that prefix so the next request reads it instead of reprocessing it.",[34,235,236,246],{},[37,237,238],{},[40,239,240,243],{},[43,241,242],{},"Cache operation",[43,244,245],{},"Price relative to input",[59,247,248,256,264],{},[40,249,250,253],{},[64,251,252],{},"5-minute cache write",[64,254,255],{},"1.25x",[40,257,258,261],{},[64,259,260],{},"1-hour cache write",[64,262,263],{},"2x",[40,265,266,268],{},[64,267,54],{},[64,269,270],{},"0.1x (0.05x on Opus 5.5, 0.025x on Fable 5.1)",[11,272,273],{},"A 5-minute cache pays for itself after a single read. On most models cached tokens also do not count toward your input rate limit, so caching raises throughput as well as cutting cost. For anything with a stable system prompt or repeated documents, it is the single biggest lever.",[227,275,277],{"id":276},"batch-processing-50-off","Batch processing: 50% off",[11,279,280],{},"The Message Batches API processes requests asynchronously at half price for both input and output. It suits anything that does not need an answer in seconds: nightly classification, document processing, evaluations, backfills. Batch and caching discounts stack.",[227,282,284],{"id":283},"tools","Tools",[197,286,287,293,304,310],{},[200,288,289,292],{},[203,290,291],{},"Web search",": $10 per 1,000 searches, plus the tokens of the results the model reads.",[200,294,295,298,299,303],{},[203,296,297],{},"Web fetch",": no extra charge, only the tokens of the fetched content. Set ",[300,301,302],"code",{},"max_content_tokens"," so one large page does not blow up a request.",[200,305,306,309],{},[203,307,308],{},"Code execution",": 1,550 free container-hours per organisation per month, then $0.05 per hour; free when used together with the current web search or web fetch tools.",[200,311,312,315],{},[203,313,314],{},"Tool definitions"," add a few hundred input tokens per request for the tool-use system prompt, plus your tool schemas. With dozens of tools or a browser toolset (about 6,600 tokens) this becomes noticeable on every call.",[227,317,319],{"id":318},"claude-managed-agents","Claude Managed Agents",[11,321,322,323,326,327,330],{},"Anthropic's hosted agent harness bills normal token prices plus ",[203,324,325],{},"$0.08 per session-hour"," while the session is actually running (idle time waiting for input or approval is free). See ",[18,328,319],{"href":329},"\u002Fen\u002Fguides\u002Fai-agents\u002Fclaude-managed-agents"," for the full picture.",[227,332,334],{"id":333},"data-residency-and-fast-mode","Data residency and fast mode",[197,336,337,351],{},[200,338,339,342,343,346,347,350],{},[203,340,341],{},"US-only inference"," (",[300,344,345],{},"inference_geo: \"us\"",") costs ",[203,348,349],{},"1.1x"," on all token categories for Claude 4.6 and later. Global routing is standard price.",[200,352,353,356],{},[203,354,355],{},"Fast mode"," (research preview) for Opus 5.5 costs $8 \u002F $40 per MTok, double the standard rate, for significantly faster output.",[26,358,360],{"id":359},"two-worked-examples","Two worked examples",[11,362,363,366],{},[203,364,365],{},"Anthropic's own example: a one-hour agent session"," on Opus 5 with 50,000 input and 15,000 output tokens costs $0.25 + $0.375 = $0.625 in tokens, plus $0.08 runtime on Managed Agents. With 40,000 of the input tokens read from cache, the token cost falls to about $0.45.",[11,368,369,372],{},[203,370,371],{},"An internal assistant on Sonnet 5.5",", our own illustration: 1,000 requests a day, each with 6,000 tokens of system prompt and reference material that is cached, 2,000 tokens of fresh input, and a 500-token answer.",[34,374,375,388],{},[37,376,377],{},[40,378,379,382,385],{},[43,380,381],{},"Line item",[43,383,384],{},"Calculation",[43,386,387],{},"Per day",[59,389,390,401,412,422],{},[40,391,392,395,398],{},[64,393,394],{},"Fresh input",[64,396,397],{},"2M tokens × $2 \u002F MTok",[64,399,400],{},"$4.00",[40,402,403,406,409],{},[64,404,405],{},"Cache reads",[64,407,408],{},"6M tokens × $0.20 \u002F MTok",[64,410,411],{},"$1.20",[40,413,414,416,419],{},[64,415,51],{},[64,417,418],{},"0.5M tokens × $10 \u002F MTok",[64,420,421],{},"$5.00",[40,423,424,429,431],{},[64,425,426],{},[203,427,428],{},"Total",[64,430],{},[64,432,433],{},[203,434,435],{},"≈ $10.20",[11,437,438],{},"That is roughly $300 a month (cache writes add a little on top). Without caching, the 6M tokens of repeated context would cost $12 a day instead of $1.20, roughly doubling the bill. On Haiku 5.5 the same workload would cost roughly a twentieth; on Opus 5.5 about double.",[11,440,441,442],{},"The pattern holds generally: ",[203,443,444],{},"output and uncached context drive the bill, not the request count.",[26,446,448],{"id":447},"spend-caps-and-billing","Spend caps and billing",[11,450,451],{},"API usage is billed separately from Claude.ai subscriptions, normally through prepaid credits in the Claude Console. Organisations sit on a usage tier, and each tier has a monthly spend cap:",[34,453,454,464],{},[37,455,456],{},[40,457,458,461],{},[43,459,460],{},"Usage tier",[43,462,463],{},"Monthly spend cap",[59,465,466,474,482],{},[40,467,468,471],{},[64,469,470],{},"Start",[64,472,473],{},"$500",[40,475,476,479],{},[64,477,478],{},"Build",[64,480,481],{},"$1,000",[40,483,484,487],{},[64,485,486],{},"Scale",[64,488,489],{},"$200,000",[11,491,492,493,496],{},"When you hit the cap, requests fail with a 429 error until the first of the next month or until you move up a tier. You can set your own lower limit on the Billing page, and ",[203,494,495],{},"per-workspace spend and rate limits"," so one team or one runaway script cannot use up everyone's budget. Rate limits (requests and tokens per minute) also depend on the tier.",[26,498,500],{"id":499},"claude-api-pricing-through-aws-google-cloud-and-azure","Claude API pricing through AWS, Google Cloud and Azure",[11,502,503],{},"Claude is also sold through Amazon Bedrock, Google Cloud, Claude Platform on AWS and Microsoft Foundry. Each bills through the cloud provider:",[197,505,506,516],{},[200,507,508,511,512,515],{},[203,509,510],{},"Bedrock and Google Cloud"," price Claude themselves. For Claude 4.5 and later models, ",[203,513,514],{},"regional and multi-region endpoints cost 10% more than global ones",". This is the route most European companies use to keep inference in an EU region.",[200,517,518,521,522,525],{},[203,519,520],{},"Claude Platform on AWS"," and ",[203,523,524],{},"Microsoft Foundry"," are operated by Anthropic and use the same per-token rates as the first-party API, billed in Claude Consumption Units ($0.01 each) through the marketplace.",[11,527,528],{},"Batch processing and some features differ by platform, so check the provider's own pricing page before you commit.",[26,530,532],{"id":531},"how-to-keep-a-companys-claude-api-bill-under-control","How to keep a company's Claude API bill under control",[11,534,535],{},"Individual developers optimise prompts. Companies need structure, because the spend is spread across teams, tools and agents nobody has a full view of.",[537,538,539,545,551,557,563,569],"ol",{},[200,540,541,544],{},[203,542,543],{},"Route by task, not by habit."," Many requests that go to Opus by default would be just as good on Sonnet or Haiku. A small evaluation set per use case tells you which.",[200,546,547,550],{},[203,548,549],{},"Cache by design."," Put stable content (instructions, tool definitions, reference documents) first in the prompt and the changing part last, so the prefix can be cached.",[200,552,553,556],{},[203,554,555],{},"Batch what can wait."," Anything overnight or bulk goes through the Batch API.",[200,558,559,562],{},[203,560,561],{},"Cap the context."," Limit how much retrieved text, tool output and conversation history goes into each request. Agents that read whole files or web pages are the usual surprise on the bill.",[200,564,565,568],{},[203,566,567],{},"Budget per team."," Separate workspaces with their own spend limits, so cost is attributable and contained.",[200,570,571,574],{},[203,572,573],{},"Watch it weekly."," Use the Console's Usage page, or the Usage and Cost API, rather than discovering the trend on the invoice.",[11,576,577,578,582,583,587],{},"Doing that across Claude Code, internal assistants and agents, and across more than one model vendor, is what a gateway is for. ",[18,579,581],{"href":580},"\u002Fen\u002Fai-hub","Walma AI Hub"," enforces budgets and model policy per team, forecasts cost per model to month end, and logs every call, running in the customer's own Azure tenant in an EU region. There is a cost calculator on the home page, and the ",[18,584,586],{"href":585},"\u002Fen\u002Fguides\u002Fclaude-code\u002Fclaude-code-pricing","Claude Code pricing guide"," covers the subscription side of the same question.",[26,589,591],{"id":590},"where-to-go-next","Where to go next",[197,593,594,601,607],{},[200,595,596,600],{},[18,597,599],{"href":598},"\u002Fen\u002Fguides\u002Fclaude-code\u002Fclaude-api-key","Claude API key",": creating keys, key types and team key management.",[200,602,603,606],{},[18,604,605],{"href":585},"Claude Code pricing",": subscriptions vs API for developers.",[200,608,609,611],{},[18,610,319],{"href":329},": what hosted agent sessions cost.",{"title":613,"searchDepth":614,"depth":615,"links":616},"",2,3,[617,618,625,626,627,628,629],{"id":28,"depth":614,"text":29},{"id":224,"depth":614,"text":225,"children":619},[620,621,622,623,624],{"id":229,"depth":615,"text":230},{"id":276,"depth":615,"text":277},{"id":283,"depth":615,"text":284},{"id":318,"depth":615,"text":319},{"id":333,"depth":615,"text":334},{"id":359,"depth":614,"text":360},{"id":447,"depth":614,"text":448},{"id":499,"depth":614,"text":500},{"id":531,"depth":614,"text":532},{"id":590,"depth":614,"text":591},"2026-10-06","Claude API prices per million input and output tokens for Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, plus prompt caching read\u002Fwrite, batch, tools and Managed Agents, worked cost examples and how to cut the bill.","md",[634,637,640,643,646],{"q":635,"a":636},"How much does the Claude API cost?","You pay per million tokens (MTok), with separate prices for input and output. At the time of writing Claude Haiku 5.5 costs $0.10 input \u002F $0.50 output per MTok (for prompts up to 100K tokens), Claude Sonnet 5.5 $2 \u002F $10, Claude Opus 5.5 $4 \u002F $20 and Claude Fable 5.1 $10 \u002F $50. Prompt caching, batch processing and data residency change the effective price; check Anthropic's pricing page for current numbers.",{"q":638,"a":639},"Is the Claude API free?","No. New Console accounts get a small amount of free credit for testing, but usage is billed per token, normally through prepaid credits. A Claude Pro, Max, Team or Enterprise subscription does not include API usage.",{"q":641,"a":642},"What is the cheapest way to use the Claude API?","Use the smallest model that does the job well (often Haiku or Sonnet), cache the repeated parts of your prompts (cache reads cost a tenth of the input price on most models), send non-urgent work through the Batch API for a 50% discount, and keep tool results and retrieved documents short. Together these often cut a bill by more than half.",{"q":644,"a":645},"Is the Claude API cheaper than a Claude subscription?","It depends on volume. For occasional use, pay-per-token is usually cheaper. For a developer using Claude Code heavily every day, a Max subscription is often cheaper than the same usage on the API. Automation, CI and products you build for others need the API regardless.",{"q":647,"a":648},"Does Claude API pricing differ in the EU?","The first-party Claude API offers global routing at standard prices or US-only inference at 1.1x; there is no EU-only option there. Through Amazon Bedrock or Google Cloud you can use regional endpoints in EU regions, which carry a 10% premium over global endpoints for Claude 4.5 and later models.",{},true,7,"\u002Fguides\u002Fclaude-code\u002Fclaude-api-pricing","10 min read",{"title":5,"description":631},"guides\u002Fclaude-code\u002Fclaude-api-pricing","claude-code","claude-api-pricing","2026-10-12","rW0Cd8gDHdpIEbMVs0eGJB_UXzbGPV2eLBJtSl8k_GE",[661,1043,1530,1794,2452,3328,4283,4773,5208],{"id":662,"title":663,"author":6,"body":664,"date":1017,"description":1018,"extension":632,"faq":1019,"meta":1035,"navigation":650,"order":1036,"path":1037,"readTime":1038,"seo":1039,"stem":1040,"topic":656,"translationId":1041,"updated":1017,"__hash__":1042},"guides\u002Fguides\u002Fclaude-code\u002Findex.md","What is Claude Code? A complete guide to Anthropic's coding agent",{"type":8,"value":665,"toc":1008},[666,669,672,676,679,721,724,728,731,794,797,816,828,832,835,861,867,871,881,887,905,919,925,929,935,941,947,951,954,960,966,972,979,983,1004],[11,667,668],{},"Claude Code is a tool you start in your terminal and give tasks in plain language: \"Find out why the login test fails and fix it.\" The agent reads the codebase, plans, changes files, runs the tests and shows you the result. Since its launch in early 2025 it has become one of the most widely used coding agents, and for many companies the first place they meet agentic AI in production.",[11,670,671],{},"This is the starting point. It covers what Claude Code can do, how it compares with the alternatives, the key concepts, and what a company needs to know about data and governance. Installation and pricing have their own pages.",[26,673,675],{"id":674},"what-claude-code-does","What Claude Code does",[11,677,678],{},"Unlike autocomplete in an editor, Claude Code works as an agent: it takes a goal and executes several steps on its own.",[197,680,681,687,693,699,705,711],{},[200,682,683,686],{},[203,684,685],{},"Understands the codebase."," Searches the project, reads relevant files, explains architecture or individual functions.",[200,688,689,692],{},[203,690,691],{},"Makes changes."," New features, refactors, bug fixes across many files, with a diff preview before every edit.",[200,694,695,698],{},[203,696,697],{},"Runs and verifies."," Tests, linters, build commands. If something fails it reads the output and corrects.",[200,700,701,704],{},[203,702,703],{},"Handles git."," Branches, commits with sensible messages, pull requests, merge conflicts.",[200,706,707,710],{},[203,708,709],{},"Connects to tools."," Through MCP servers it reaches GitHub, Jira, databases, browsers and internal systems.",[200,712,713,716,717,720],{},[203,714,715],{},"Loads know-how."," Through skills and a ",[300,718,719],{},"CLAUDE.md"," file in the repository it knows your team's conventions.",[11,722,723],{},"It runs in the terminal, as an extension in VS Code and JetBrains, as a GitHub Action in CI, and through an SDK inside your own agents.",[26,725,727],{"id":726},"installation-in-brief","Installation in brief",[11,729,730],{},"The fastest route is the native installer, which needs no Node.js:",[732,733,737],"pre",{"className":734,"code":735,"language":736,"meta":613,"style":613},"language-bash shiki shiki-themes github-dark","# macOS and Linux\ncurl -fsSL https:\u002F\u002Fclaude.ai\u002Finstall.sh | bash\n\n# Windows (PowerShell)\nirm https:\u002F\u002Fclaude.ai\u002Finstall.ps1 | iex\n","bash",[300,738,739,748,769,774,780],{"__ignoreMap":613},[740,741,744],"span",{"class":742,"line":743},"line",1,[740,745,747],{"class":746},"sAwPA","# macOS and Linux\n",[740,749,750,754,758,762,766],{"class":742,"line":614},[740,751,753],{"class":752},"svObZ","curl",[740,755,757],{"class":756},"sDLfK"," -fsSL",[740,759,761],{"class":760},"sU2Wk"," https:\u002F\u002Fclaude.ai\u002Finstall.sh",[740,763,765],{"class":764},"snl16"," |",[740,767,768],{"class":752}," bash\n",[740,770,771],{"class":742,"line":615},[740,772,773],{"emptyLinePlaceholder":650},"\n",[740,775,777],{"class":742,"line":776},4,[740,778,779],{"class":746},"# Windows (PowerShell)\n",[740,781,783,786,789,791],{"class":742,"line":782},5,[740,784,785],{"class":752},"irm",[740,787,788],{"class":760}," https:\u002F\u002Fclaude.ai\u002Finstall.ps1",[740,790,765],{"class":764},[740,792,793],{"class":752}," iex\n",[11,795,796],{},"Or via npm if Node.js 18 or later is present:",[732,798,800],{"className":734,"code":799,"language":736,"meta":613,"style":613},"npm install -g @anthropic-ai\u002Fclaude-code\n",[300,801,802],{"__ignoreMap":613},[740,803,804,807,810,813],{"class":742,"line":743},[740,805,806],{"class":752},"npm",[740,808,809],{"class":760}," install",[740,811,812],{"class":756}," -g",[740,814,815],{"class":760}," @anthropic-ai\u002Fclaude-code\n",[11,817,818,819,822,823,827],{},"Then run ",[300,820,821],{},"claude"," in a project folder and sign in on first launch. Step-by-step instructions, common errors and enterprise setups are in ",[18,824,826],{"href":825},"\u002Fen\u002Fguides\u002Fclaude-code\u002Fhow-to-install-claude-code","How to install Claude Code",".",[26,829,831],{"id":830},"what-claude-code-costs","What Claude Code costs",[11,833,834],{},"Claude Code is not sold separately. It is included in Anthropic's subscriptions or billed per token through the API:",[197,836,837,843,849,855],{},[200,838,839,842],{},[203,840,841],{},"Pro",": the entry point for individuals, with a usage allowance.",[200,844,845,848],{},[203,846,847],{},"Max",": much higher allowances in two tiers, for developers who use the agent all day.",[200,850,851,854],{},[203,852,853],{},"Team and Enterprise",": seats with central administration; Claude Code is in the premium seats.",[200,856,857,860],{},[203,858,859],{},"API",": per-token billing, right for CI pipelines and custom agents.",[11,862,863,864,827],{},"The amounts, what \"allowance\" means in practice, and how to control cost across a team are in ",[18,865,866],{"href":585},"Is Claude Code free? Pricing explained",[26,868,870],{"id":869},"the-key-concepts","The key concepts",[11,872,873,876,877,880],{},[203,874,875],{},"CLAUDE.md."," A Markdown file in the repository, loaded at the start of every session. Build and test commands, conventions, and anything the agent must always know. ",[300,878,879],{},"\u002Finit"," drafts one from what it finds in the repo.",[11,882,883,886],{},[203,884,885],{},"Permissions."," Claude Code asks before it edits files or runs commands. You can allow specific commands permanently, keep the agent in a read-only plan mode, or let it run more freely in trusted environments such as a sandbox or CI.",[11,888,889,892,893,896,897,900,901,827],{},[203,890,891],{},"MCP servers."," Through the Model Context Protocol the agent gets access to external systems. Add one with ",[300,894,895],{},"claude mcp add","; inside a session ",[300,898,899],{},"\u002Fmcp"," shows status and starts authentication. See ",[18,902,904],{"href":903},"\u002Fen\u002Fguides\u002Fmcp\u002Fwhat-is-an-mcp-server","What is an MCP server?",[11,906,907,910,911,914,915,827],{},[203,908,909],{},"Skills."," Folders with a ",[300,912,913],{},"SKILL.md"," that teach the agent how a specific task is done in your team, loaded only when relevant. See ",[18,916,918],{"href":917},"\u002Fen\u002Fguides\u002Fskills","Claude skills explained",[11,920,921,924],{},[203,922,923],{},"Hooks, subagents, plugins."," Hooks run your own scripts before or after certain actions (a formatter after every edit, say). Subagents handle subtasks in their own context. Plugins bundle skills, commands, hooks and MCP servers for distribution to a team.",[26,926,928],{"id":927},"claude-code-compared","Claude Code compared",[11,930,931,934],{},[203,932,933],{},"Claude Code vs Cursor."," Cursor is an editor with AI built in. Claude Code is an agent that plugs into any editor and into CI. Cursor is stronger for interactive editing, Claude Code for longer multi-step tasks. Many teams use both, often with Claude models in both.",[11,936,937,940],{},[203,938,939],{},"Claude Code vs Codex."," OpenAI's Codex is the direct counterpart with GPT models: CLI, editor integration, cloud execution. The choice usually comes down to which models the company has approved. A gateway that serves both takes the heat out of the decision.",[11,942,943,946],{},[203,944,945],{},"Claude Code vs GitHub Copilot."," Copilot started as autocomplete and now has an agent mode with model choice, including Claude. It is tightly tied to GitHub. Claude Code is more ecosystem-independent and more deeply built for agentic work.",[26,948,950],{"id":949},"data-protection-and-running-it-in-the-eu","Data protection and running it in the EU",[11,952,953],{},"This is where most European rollouts stall, and it is solvable.",[11,955,956,959],{},[203,957,958],{},"Default operation."," With an Anthropic account, requests go through Anthropic's API in the US. Team, Enterprise and API usage fall under the commercial terms, under which data is not used for training, and a data processing agreement is available.",[11,961,962,965],{},[203,963,964],{},"EU regions."," Claude Code supports Amazon Bedrock, Google Vertex AI and Microsoft Foundry as backends, so model calls can run in an EU region such as Frankfurt or Sweden Central. The switch is a few environment variables; developers keep the same tool.",[11,967,968,971],{},[203,969,970],{},"What a gateway adds."," Even with an EU region, questions remain: who may use which models, with what budget? Which MCP servers are approved? Where do credentials live? What did the agent do yesterday at 14:32? A gateway between developer and model answers all of them: policy, budgets, approved servers and a full log in one place. Claude Code supports a configurable base URL, so every request can go through the gateway without the developer noticing.",[11,973,974,975,978],{},"Walma AI Hub is that gateway, running in the customer's own Azure tenant in an EU region, for Claude Code, Codex, Cursor and the MCP servers the company approves. Developers install one signed client and work with one key; there is no client-side switch to turn policy off. If you are rolling Claude Code out beyond a handful of developers, a ",[18,976,977],{"href":580},"20-minute walkthrough"," is the fastest way to settle the open questions.",[26,980,982],{"id":981},"further-reading","Further reading",[197,984,985,990,994,999],{},[200,986,987,989],{},[18,988,826],{"href":825}," on macOS, Linux and Windows.",[200,991,992,827],{},[18,993,866],{"href":585},[200,995,996,998],{},[18,997,904],{"href":903}," How the agent reaches tools and data.",[200,1000,1001,1003],{},[18,1002,918],{"href":917},". Making team know-how reusable.",[1005,1006,1007],"style",{},"html pre.shiki code .sAwPA, html code.shiki .sAwPA{--shiki-default:#6A737D}html pre.shiki code .svObZ, html code.shiki .svObZ{--shiki-default:#B392F0}html pre.shiki code .sDLfK, html code.shiki .sDLfK{--shiki-default:#79B8FF}html pre.shiki code .sU2Wk, html code.shiki .sU2Wk{--shiki-default:#9ECBFF}html pre.shiki code .snl16, html code.shiki .snl16{--shiki-default:#F97583}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}",{"title":613,"searchDepth":614,"depth":615,"links":1009},[1010,1011,1012,1013,1014,1015,1016],{"id":674,"depth":614,"text":675},{"id":726,"depth":614,"text":727},{"id":830,"depth":614,"text":831},{"id":869,"depth":614,"text":870},{"id":927,"depth":614,"text":928},{"id":949,"depth":614,"text":950},{"id":981,"depth":614,"text":982},"2026-09-11","Claude Code is Anthropic's agentic coding tool that runs in your terminal, reads your codebase, edits files, runs tests and opens pull requests. This guide covers what it does, how it compares with Cursor, Codex and Copilot, how to install it, what it costs, and how to run it in an EU environment with policy and logging.",[1020,1023,1026,1029,1032],{"q":1021,"a":1022},"What is Claude Code?","Claude Code is a command-line tool from Anthropic that runs Claude as a coding agent in your terminal. It reads your codebase, plans changes, edits files, runs tests and commands, and creates commits and pull requests, in a conversation with you.",{"q":1024,"a":1025},"Is Claude Code free?","No. Claude Code is included in the Pro, Max, Team and Enterprise plans, or billed per token through the API. There is no permanently free tier, though new API accounts sometimes get trial credits.",{"q":1027,"a":1028},"Does Claude Code work on Windows?","Yes. Claude Code runs natively on Windows 10 and 11 (with Git for Windows), on macOS and Linux, and inside WSL. Install it with the native installer script or via npm.",{"q":1030,"a":1031},"Claude Code vs Cursor: which is better?","They are different tools. Cursor is an editor with an AI assistant built in; Claude Code is an agent that works from the terminal, independent of your editor, and can be embedded in VS Code, JetBrains and CI. Cursor is stronger for interactive editing, Claude Code for longer multi-step tasks. Many teams use both.",{"q":1033,"a":1034},"Where is my code processed when I use Claude Code?","By default at Anthropic in the US. Companies in the EU can run Claude Code through Amazon Bedrock, Google Vertex AI or Microsoft Foundry in an EU region, or through a gateway such as Walma AI Hub in their own Azure tenant.",{},0,"\u002Fguides\u002Fclaude-code","12 min read",{"title":663,"description":1018},"guides\u002Fclaude-code\u002Findex","claude-code-guide","_2hCTbI-2-EcxQYApmXsrcjnQhUGQJgYBYjJC9880eM",{"id":1044,"title":1045,"author":6,"body":1046,"date":1017,"description":1506,"extension":632,"faq":1507,"meta":1523,"navigation":650,"order":743,"path":1524,"readTime":1525,"seo":1526,"stem":1527,"topic":656,"translationId":1528,"updated":1017,"__hash__":1529},"guides\u002Fguides\u002Fclaude-code\u002Fhow-to-install-claude-code.md","How to install Claude Code on macOS, Linux and Windows",{"type":8,"value":1047,"toc":1495},[1048,1058,1062,1098,1101,1105,1108,1129,1164,1167,1179,1183,1186,1200,1207,1211,1214,1232,1235,1238,1257,1261,1267,1275,1278,1281,1344,1348,1351,1371,1375,1387,1393,1403,1411,1415,1418,1447,1450,1454,1457,1486,1492],[11,1049,1050,1051,1053,1054,827],{},"Installing Claude Code takes under two minutes. The problems come afterwards: not signed in, wrong project folder, no ",[300,1052,719],{},", or the question of where the data actually goes. This guide walks through everything in order. For what Claude Code is and does, start with the ",[18,1055,1057],{"href":1056},"\u002Fen\u002Fguides\u002Fclaude-code","overview",[26,1059,1061],{"id":1060},"requirements","Requirements",[197,1063,1064,1078,1085,1088],{},[200,1065,1066,1069,1070,1073,1074,1077],{},[203,1067,1068],{},"macOS"," 10.15 or later, ",[203,1071,1072],{},"Linux"," (Ubuntu 20.04+, Debian 10+ or equivalent) or ",[203,1075,1076],{},"Windows"," 10\u002F11.",[200,1079,1080,1081,1084],{},"On Windows: ",[203,1082,1083],{},"Git for Windows",", because Claude Code uses its Bash for shell commands. WSL 1 or 2 also works.",[200,1086,1087],{},"A terminal and internet access.",[200,1089,1090,1091,1094,1095,1097],{},"An Anthropic account (Pro, Max, Team or Enterprise) ",[203,1092,1093],{},"or"," an API key ",[203,1096,1093],{}," access to Bedrock, Vertex AI or Microsoft Foundry.",[11,1099,1100],{},"Node.js is only needed for the npm route.",[26,1102,1104],{"id":1103},"option-1-native-installer-recommended","Option 1: Native installer (recommended)",[11,1106,1107],{},"The native installer places a self-contained binary and keeps it up to date automatically.",[732,1109,1111],{"className":734,"code":1110,"language":736,"meta":613,"style":613},"# macOS and Linux\ncurl -fsSL https:\u002F\u002Fclaude.ai\u002Finstall.sh | bash\n",[300,1112,1113,1117],{"__ignoreMap":613},[740,1114,1115],{"class":742,"line":743},[740,1116,747],{"class":746},[740,1118,1119,1121,1123,1125,1127],{"class":742,"line":614},[740,1120,753],{"class":752},[740,1122,757],{"class":756},[740,1124,761],{"class":760},[740,1126,765],{"class":764},[740,1128,768],{"class":752},[732,1130,1134],{"className":1131,"code":1132,"language":1133,"meta":613,"style":613},"language-powershell shiki shiki-themes github-dark","# Windows, PowerShell\nirm https:\u002F\u002Fclaude.ai\u002Finstall.ps1 | iex\n","powershell",[300,1135,1136,1141],{"__ignoreMap":613},[740,1137,1138],{"class":742,"line":743},[740,1139,1140],{"class":746},"# Windows, PowerShell\n",[740,1142,1143,1147,1150,1153,1156,1159,1162],{"class":742,"line":614},[740,1144,1146],{"class":1145},"s95oV","irm https:",[740,1148,1149],{"class":764},"\u002F\u002F",[740,1151,1152],{"class":1145},"claude.ai",[740,1154,1155],{"class":764},"\u002F",[740,1157,1158],{"class":1145},"install.ps1 ",[740,1160,1161],{"class":764},"|",[740,1163,793],{"class":1145},[11,1165,1166],{},"Open a new terminal and check:",[732,1168,1170],{"className":734,"code":1169,"language":736,"meta":613,"style":613},"claude --version\n",[300,1171,1172],{"__ignoreMap":613},[740,1173,1174,1176],{"class":742,"line":743},[740,1175,821],{"class":752},[740,1177,1178],{"class":756}," --version\n",[26,1180,1182],{"id":1181},"option-2-npm","Option 2: npm",[11,1184,1185],{},"If Node.js 18 or later is installed:",[732,1187,1188],{"className":734,"code":799,"language":736,"meta":613,"style":613},[300,1189,1190],{"__ignoreMap":613},[740,1191,1192,1194,1196,1198],{"class":742,"line":743},[740,1193,806],{"class":752},[740,1195,809],{"class":760},[740,1197,812],{"class":756},[740,1199,815],{"class":760},[11,1201,1202,1203,1206],{},"Do not use ",[300,1204,1205],{},"sudo",". If npm cannot install global packages without root, configure a user-level prefix or use the native installer instead.",[26,1208,1210],{"id":1209},"first-sign-in","First sign-in",[11,1212,1213],{},"Change into a project folder and start the agent:",[732,1215,1217],{"className":734,"code":1216,"language":736,"meta":613,"style":613},"cd ~\u002Fprojects\u002Fmy-repo\nclaude\n",[300,1218,1219,1227],{"__ignoreMap":613},[740,1220,1221,1224],{"class":742,"line":743},[740,1222,1223],{"class":756},"cd",[740,1225,1226],{"class":760}," ~\u002Fprojects\u002Fmy-repo\n",[740,1228,1229],{"class":742,"line":614},[740,1230,1231],{"class":752},"claude\n",[11,1233,1234],{},"On first launch a browser window opens for sign-in. You can use a Claude account (Pro, Max, Team, Enterprise) or an Anthropic Console account (API billing). After that the agent is ready.",[11,1236,1237],{},"For scripts, CI and environments without a browser, set a key instead:",[732,1239,1241],{"className":734,"code":1240,"language":736,"meta":613,"style":613},"export ANTHROPIC_API_KEY=sk-ant-...\n",[300,1242,1243],{"__ignoreMap":613},[740,1244,1245,1248,1251,1254],{"class":742,"line":743},[740,1246,1247],{"class":764},"export",[740,1249,1250],{"class":1145}," ANTHROPIC_API_KEY",[740,1252,1253],{"class":764},"=",[740,1255,1256],{"class":1145},"sk-ant-...\n",[26,1258,1260],{"id":1259},"setting-up-the-project","Setting up the project",[11,1262,1263,1264,1266],{},"The most important step after installation is a ",[300,1265,719],{}," in the repository. It is loaded every session and should say how the project is built and tested, which conventions apply, and what the agent must not touch.",[732,1268,1273],{"className":1269,"code":1271,"language":1272},[1270],"language-text","> \u002Finit\n","text",[300,1274,1271],{"__ignoreMap":613},[11,1276,1277],{},"drafts one from what the agent finds in the repo. Edit it and commit it; it then applies to the whole team.",[11,1279,1280],{},"Other useful commands inside a session:",[34,1282,1283,1293],{},[37,1284,1285],{},[40,1286,1287,1290],{},[43,1288,1289],{},"Command",[43,1291,1292],{},"Purpose",[59,1294,1295,1305,1315,1324,1334],{},[40,1296,1297,1302],{},[64,1298,1299],{},[300,1300,1301],{},"\u002Fhelp",[64,1303,1304],{},"All commands",[40,1306,1307,1312],{},[64,1308,1309],{},[300,1310,1311],{},"\u002Fstatus",[64,1313,1314],{},"Version, account, model, context",[40,1316,1317,1321],{},[64,1318,1319],{},[300,1320,899],{},[64,1322,1323],{},"Show and authenticate MCP servers",[40,1325,1326,1331],{},[64,1327,1328],{},[300,1329,1330],{},"\u002Fpermissions",[64,1332,1333],{},"Allowed and blocked actions",[40,1335,1336,1341],{},[64,1337,1338],{},[300,1339,1340],{},"\u002Fclear",[64,1342,1343],{},"Clear the context",[26,1345,1347],{"id":1346},"editor-integration","Editor integration",[11,1349,1350],{},"Claude Code is editor-independent but integrates with the common ones:",[197,1352,1353,1359,1365],{},[200,1354,1355,1358],{},[203,1356,1357],{},"VS Code",": install the extension from the marketplace; diffs appear in the editor.",[200,1360,1361,1364],{},[203,1362,1363],{},"JetBrains"," (IntelliJ, PyCharm, WebStorm and the rest): plugin from the marketplace.",[200,1366,1367,1370],{},[203,1368,1369],{},"Any editor's terminal",": works everywhere, no plugin.",[26,1372,1374],{"id":1373},"common-errors","Common errors",[11,1376,1377,1382,1383,1386],{},[203,1378,1379,827],{},[300,1380,1381],{},"claude: command not found"," Restart the terminal so the PATH is refreshed. With npm, check that the global bin directory is on the PATH (",[300,1384,1385],{},"npm bin -g",").",[11,1388,1389,1392],{},[203,1390,1391],{},"Windows: shell commands fail."," Git for Windows is missing or not on the PATH. Restart the terminal after installing Git.",[11,1394,1395,1398,1399,1402],{},[203,1396,1397],{},"Sign-in fails on a corporate network."," Set proxy variables (",[300,1400,1401],{},"HTTPS_PROXY","); with TLS inspection, register the corporate certificate with Node or the system store.",[11,1404,1405,1408,1409,827],{},[203,1406,1407],{},"\"Rate limit\" or allowance exhausted."," Pro and Max have usage windows; API accounts have account limits. See ",[18,1410,605],{"href":585},[26,1412,1414],{"id":1413},"running-through-an-eu-region","Running through an EU region",[11,1416,1417],{},"By default, requests go to Anthropic in the US. If your data protection rules require EU processing, run Claude Code through one of the supported cloud providers. The switch is environment variables, for example for Bedrock:",[732,1419,1421],{"className":734,"code":1420,"language":736,"meta":613,"style":613},"export CLAUDE_CODE_USE_BEDROCK=1\nexport AWS_REGION=eu-central-1\n",[300,1422,1423,1435],{"__ignoreMap":613},[740,1424,1425,1427,1430,1432],{"class":742,"line":743},[740,1426,1247],{"class":764},[740,1428,1429],{"class":1145}," CLAUDE_CODE_USE_BEDROCK",[740,1431,1253],{"class":764},[740,1433,1434],{"class":756},"1\n",[740,1436,1437,1439,1442,1444],{"class":742,"line":614},[740,1438,1247],{"class":764},[740,1440,1441],{"class":1145}," AWS_REGION",[740,1443,1253],{"class":764},[740,1445,1446],{"class":1145},"eu-central-1\n",[11,1448,1449],{},"Equivalent variables exist for Vertex AI and Microsoft Foundry. Developers keep using the same tool; only the backend changes.",[26,1451,1453],{"id":1452},"running-through-a-gateway","Running through a gateway",[11,1455,1456],{},"For more than a handful of developers, the region alone is not enough. You need one place for model approvals, budgets, allowed MCP servers and the log. Claude Code supports a configurable base URL so that every request goes through a gateway:",[732,1458,1460],{"className":734,"code":1459,"language":736,"meta":613,"style":613},"export ANTHROPIC_BASE_URL=https:\u002F\u002Fhub.your-company.eu\nexport ANTHROPIC_AUTH_TOKEN=...\n",[300,1461,1462,1474],{"__ignoreMap":613},[740,1463,1464,1466,1469,1471],{"class":742,"line":743},[740,1465,1247],{"class":764},[740,1467,1468],{"class":1145}," ANTHROPIC_BASE_URL",[740,1470,1253],{"class":764},[740,1472,1473],{"class":1145},"https:\u002F\u002Fhub.your-company.eu\n",[740,1475,1476,1478,1481,1483],{"class":742,"line":614},[740,1477,1247],{"class":764},[740,1479,1480],{"class":1145}," ANTHROPIC_AUTH_TOKEN",[740,1482,1253],{"class":764},[740,1484,1485],{"class":1145},"...\n",[11,1487,1488,1489,827],{},"Walma AI Hub is such a gateway, running in your own Azure tenant in an EU region. Developers install a signed client for Windows and macOS that sets this configuration, and work with one key for Claude Code, Codex, Cursor and every approved MCP server. Policy is enforced in the gateway, not on the laptop. ",[18,1490,1491],{"href":580},"More about AI Hub",[1005,1493,1494],{},"html pre.shiki code .sAwPA, html code.shiki .sAwPA{--shiki-default:#6A737D}html pre.shiki code .svObZ, html code.shiki .svObZ{--shiki-default:#B392F0}html pre.shiki code .sDLfK, html code.shiki .sDLfK{--shiki-default:#79B8FF}html pre.shiki code .sU2Wk, html code.shiki .sU2Wk{--shiki-default:#9ECBFF}html pre.shiki code .snl16, html code.shiki .snl16{--shiki-default:#F97583}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html pre.shiki code .s95oV, html code.shiki .s95oV{--shiki-default:#E1E4E8}",{"title":613,"searchDepth":614,"depth":615,"links":1496},[1497,1498,1499,1500,1501,1502,1503,1504,1505],{"id":1060,"depth":614,"text":1061},{"id":1103,"depth":614,"text":1104},{"id":1181,"depth":614,"text":1182},{"id":1209,"depth":614,"text":1210},{"id":1259,"depth":614,"text":1260},{"id":1346,"depth":614,"text":1347},{"id":1373,"depth":614,"text":1374},{"id":1413,"depth":614,"text":1414},{"id":1452,"depth":614,"text":1453},"Step-by-step installation of Claude Code with the native installer or npm, first sign-in, project setup with CLAUDE.md, the VS Code and JetBrains integrations, common errors, and how to run it through an EU region or a gateway.",[1508,1511,1514,1517,1520],{"q":1509,"a":1510},"Do I need Node.js to install Claude Code?","Not any more. The native installer is self-contained. Only the npm route requires Node.js 18 or later.",{"q":1512,"a":1513},"How do I update Claude Code?","The native installer updates itself in the background. With npm, run npm install -g @anthropic-ai\u002Fclaude-code again. Inside a session, \u002Fstatus shows the current version.",{"q":1515,"a":1516},"Claude Code won't start on Windows. What should I check?","Make sure Git for Windows is installed; Claude Code uses its Bash for shell commands. Restart the terminal after installing so the PATH is refreshed. As an alternative, Claude Code runs reliably inside WSL.",{"q":1518,"a":1519},"Can I use Claude Code without an Anthropic account?","Yes, with an API key via the ANTHROPIC_API_KEY environment variable, or through Amazon Bedrock, Google Vertex AI or Microsoft Foundry using those clouds' credentials.",{"q":1521,"a":1522},"How do I use Claude Code in VS Code?","Install the Claude Code extension from the VS Code marketplace; diffs then appear in the editor and you can start sessions from the sidebar. Running the CLI in VS Code's integrated terminal works too, with no extension.",{},"\u002Fguides\u002Fclaude-code\u002Fhow-to-install-claude-code","8 min read",{"title":1045,"description":1506},"guides\u002Fclaude-code\u002Fhow-to-install-claude-code","claude-code-install","YGHP6DTDshBdF9TM0FMpxWREc6MWSDQvZG6vgvmFWzU",{"id":1531,"title":1532,"author":6,"body":1533,"date":1017,"description":1771,"extension":632,"faq":1772,"meta":1787,"navigation":650,"order":614,"path":1788,"readTime":1789,"seo":1790,"stem":1791,"topic":656,"translationId":1792,"updated":658,"__hash__":1793},"guides\u002Fguides\u002Fclaude-code\u002Fclaude-code-pricing.md","Is Claude Code free? Claude Code pricing and cost control explained",{"type":8,"value":1534,"toc":1764},[1535,1540,1543,1547,1649,1656,1660,1663,1666,1670,1680,1683,1703,1706,1710,1716,1721,1724,1728,1731,1757],[11,1536,1537,1538,827],{},"Claude Code has no price of its own. It is a tool that runs on one of Anthropic's subscriptions or on pay-per-token API billing. What you pay therefore depends on which route you choose, which model you use, and how hard the agent works. This guide explains the options, puts the numbers in context, and shows how to keep costs under control across a team. For what Claude Code is, start with the ",[18,1539,1057],{"href":1056},[11,1541,1542],{},"A note up front: prices change. The figures here are as of September 2026 in US dollars; Anthropic's pricing page has the current amounts and local currency.",[26,1544,1546],{"id":1545},"the-options-at-a-glance","The options at a glance",[34,1548,1549,1565],{},[37,1550,1551],{},[40,1552,1553,1556,1559,1562],{},[43,1554,1555],{},"Option",[43,1557,1558],{},"Who it is for",[43,1560,1561],{},"Billing",[43,1563,1564],{},"Claude Code included",[59,1566,1567,1580,1594,1608,1622,1636],{},[40,1568,1569,1571,1574,1577],{},[64,1570,841],{},[64,1572,1573],{},"Individuals, occasional use",[64,1575,1576],{},"~20 USD\u002Fmonth",[64,1578,1579],{},"Yes, with a usage window",[40,1581,1582,1585,1588,1591],{},[64,1583,1584],{},"Max 5x",[64,1586,1587],{},"Developers using it daily",[64,1589,1590],{},"~100 USD\u002Fmonth",[64,1592,1593],{},"Yes, ~5× the Pro allowance",[40,1595,1596,1599,1602,1605],{},[64,1597,1598],{},"Max 20x",[64,1600,1601],{},"Heavy users, parallel sessions",[64,1603,1604],{},"~200 USD\u002Fmonth",[64,1606,1607],{},"Yes, ~20× the Pro allowance",[40,1609,1610,1613,1616,1619],{},[64,1611,1612],{},"Team",[64,1614,1615],{},"Teams of 2 to 150, central admin",[64,1617,1618],{},"$25 Standard \u002F $125 Premium per seat (annual: $20 \u002F $100)",[64,1620,1621],{},"Yes, on every seat",[40,1623,1624,1627,1630,1633],{},[64,1625,1626],{},"Enterprise",[64,1628,1629],{},"20+ seats, SCIM, audit logs, contracts",[64,1631,1632],{},"$20 per seat (annual) + usage at API rates",[64,1634,1635],{},"Yes",[40,1637,1638,1640,1643,1646],{},[64,1639,859],{},[64,1641,1642],{},"CI, custom agents, full control",[64,1644,1645],{},"Per token",[64,1647,1648],{},"Yes, via API key",[11,1650,1651,1652,827],{},"For every plan side by side, including which models and features each one gets, see ",[18,1653,1655],{"href":1654},"\u002Fen\u002Fguides\u002Fai-tools\u002Fclaude-plans","Claude plans compared",[26,1657,1659],{"id":1658},"what-usage-allowance-means","What \"usage allowance\" means",[11,1661,1662],{},"Pro and Max do not bill per token. They give you an allowance within rolling time windows. When it is used up, the agent pauses until the next window opens. There are additional weekly caps for very heavy users.",[11,1664,1665],{},"How far the allowance stretches depends on the model and the kind of work. Short, targeted tasks in a small repository use little. Long sessions where the agent reads large files, reruns tests and calls many tools use a lot. Rule of thumb: Pro covers a few hours a day with breaks, Max 5x a full working day, Max 20x parallel sessions or very large codebases.",[26,1667,1669],{"id":1668},"api-billing-what-a-developer-spends","API billing: what a developer spends",[11,1671,1672,1673,1677,1678,827],{},"Through the API you pay per input and output token, with different prices per model. Sonnet-class models are considerably cheaper than Opus-class ones, and prompt caching cuts the cost of repeated inputs substantially; Claude Code uses it automatically. The current per-model prices, caching and batch discounts are in our ",[18,1674,1676],{"href":1675},"\u002Fen\u002Fguides\u002Fclaude-code\u002Fclaude-api-pricing","Claude API pricing guide",", and getting a key is covered in ",[18,1679,599],{"href":598},[11,1681,1682],{},"From the rollouts we support, the orders of magnitude are:",[197,1684,1685,1691,1697],{},[200,1686,1687,1690],{},[203,1688,1689],{},"Occasional use"," (a few tasks a week): single to low double digits in dollars per month.",[200,1692,1693,1696],{},[203,1694,1695],{},"Daily use"," (several hours): mid double digits to low triple digits.",[200,1698,1699,1702],{},[203,1700,1701],{},"Heavy use with Opus models and large contexts",": several hundred dollars is possible.",[11,1704,1705],{},"The point that surprises finance teams: the spread between developers is wide, and a single session can cost many times the daily average. Without budget limits you find out on the monthly invoice.",[26,1707,1709],{"id":1708},"subscription-or-api","Subscription or API?",[11,1711,1712,1715],{},[203,1713,1714],{},"Subscription"," when developers use the tool interactively and you want predictable per-head cost. Team and Enterprise add central administration, SSO and the commercial terms under which data is not used for training.",[11,1717,1718,1720],{},[203,1719,859],{}," when Claude Code runs in CI pipelines, automations or your own agents, when you want to choose the backend yourself (an EU region via Bedrock, Vertex AI or Microsoft Foundry), or when you operate a gateway with your own budgets.",[11,1722,1723],{},"Many companies combine the two: subscriptions for interactive use, API for automation.",[26,1725,1727],{"id":1726},"controlling-cost-across-a-team","Controlling cost across a team",[11,1729,1730],{},"Four measures that make the difference in practice:",[537,1732,1733,1739,1745,1751],{},[200,1734,1735,1738],{},[203,1736,1737],{},"Budgets and hard caps"," per user, team and project. Not a report after the fact, but a limit the agent cannot exceed.",[200,1740,1741,1744],{},[203,1742,1743],{},"Model policies."," Not every task needs the most expensive model. A policy that defaults to a cheaper model and unlocks the expensive one on request often halves cost with no noticeable quality loss.",[200,1746,1747,1750],{},[203,1748,1749],{},"A forecast to month end"," per model and team, based on actual usage over the last weeks rather than a straight line.",[200,1752,1753,1756],{},[203,1754,1755],{},"A log per call",": who, when, which model, how many tokens, which repository. Without it every cost discussion is a guess.",[11,1758,1759,1760,1763],{},"These four are why companies put a gateway in front of Claude Code. Walma AI Hub enforces budgets and model policy in the gateway, forecasts cost per model to month end, and logs every call, for Claude Code, Codex, Cursor and the approved MCP servers, running in the customer's own Azure tenant in an EU region. There is a cost calculator on the home page; for a concrete scenario, a ",[18,1761,1762],{"href":580},"walkthrough with an engineer"," is the fastest route.",{"title":613,"searchDepth":614,"depth":615,"links":1765},[1766,1767,1768,1769,1770],{"id":1545,"depth":614,"text":1546},{"id":1658,"depth":614,"text":1659},{"id":1668,"depth":614,"text":1669},{"id":1708,"depth":614,"text":1709},{"id":1726,"depth":614,"text":1727},"What Claude Code costs: the differences between Pro, Max, Team, Enterprise and API billing, what a usage allowance means in practice, how much a developer typically spends per month, and how to keep costs under control across a team.",[1773,1775,1778,1781,1784],{"q":1024,"a":1774},"No. There is no permanently free tier for Claude Code. It is included in the Pro, Max, Team and Enterprise subscriptions, or billed per token through the API. New API accounts sometimes receive a small trial credit.",{"q":1776,"a":1777},"How much does Claude Code cost per month?","Through a subscription: Pro is around 20 US dollars per month, Max around 100 or 200 depending on tier, Team costs $25 or $125 per seat, and Enterprise $20 per seat plus usage at API rates. Through the API you pay per token, which for a developer using the agent daily typically lands in the tens to low hundreds of dollars per month. Check Anthropic's pricing page for current amounts.",{"q":1779,"a":1780},"Is Claude Pro enough for Claude Code?","For occasional use, yes. Developers who run the agent for several hours a day regularly hit Pro's usage window and are better served by Max.",{"q":1782,"a":1783},"Is Claude Code cheaper through the API or a subscription?","For light use the API is cheaper; for heavy daily use a Max subscription usually is. Many companies combine subscriptions for interactive work with API billing for CI and automation.",{"q":1785,"a":1786},"How do we control Claude Code costs across a team?","Budgets and hard caps per user and team, model policies that default to cheaper models, a forecast to month end, and a log per call. All of that is easiest to enforce in a gateway that every request passes through.",{},"\u002Fguides\u002Fclaude-code\u002Fclaude-code-pricing","9 min read",{"title":1532,"description":1771},"guides\u002Fclaude-code\u002Fclaude-code-pricing","claude-code-pricing","SAmQLu1sp57levLIu29Cgp54vK12vjIIwotevxyt7nc",{"id":1795,"title":1796,"author":6,"body":1797,"date":2428,"description":2429,"extension":632,"faq":2430,"meta":2446,"navigation":650,"order":615,"path":2447,"readTime":1525,"seo":2448,"stem":2449,"topic":656,"translationId":2450,"updated":2428,"__hash__":2451},"guides\u002Fguides\u002Fclaude-code\u002Fcursor-vs-claude-code.md","Cursor vs Claude Code: which AI coding tool fits your team?",{"type":8,"value":1798,"toc":2417},[1799,1805,1809,1823,1826,1830,1839,1858,1865,1870,1898,1907,1910,1914,1976,1979,1982,1986,1989,1993,2023,2027,2063,2093,2096,2100,2103,2172,2178,2182,2185,2212,2262,2284,2303,2307,2317,2334,2337,2341,2387,2391,2402,2414],[11,1800,1801,1802,1804],{},"\"Cursor vs Claude Code\" is one of the most searched questions in AI coding right now, and most answers are out of date within a quarter. Both tools have converged a lot: Cursor now has a terminal agent and cloud agents, Claude Code now has IDE extensions and a desktop app. The difference that remains is where each one starts from. This guide compares them as of October 2026, based on the vendors' own docs and pricing pages. If you are new to Claude Code, read the ",[18,1803,1057],{"href":1056}," first.",[26,1806,1808],{"id":1807},"the-short-version","The short version",[197,1810,1811,1817],{},[200,1812,1813,1816],{},[203,1814,1815],{},"Cursor"," is an editor with AI built into every surface: Tab autocomplete as you type, an agent panel, cloud agents, and a choice of models from several labs.",[200,1818,1819,1822],{},[203,1820,1821],{},"Claude Code"," is an agent with several front ends: the CLI, VS Code and JetBrains extensions, a desktop app, the web, Slack and CI. It runs Claude models only.",[11,1824,1825],{},"If your developers live in the editor and want fast inline help with occasional agent tasks, Cursor fits more naturally. If you want an agent that does long multi-step work, scripts into CI and behaves the same across every surface, Claude Code fits better. Many teams end up using both, and that works fine (see the last section).",[26,1827,1829],{"id":1828},"interface-ide-first-vs-agent-first","Interface: IDE first vs agent first",[11,1831,1832,1834,1835,1838],{},[203,1833,1815],{}," is a fork of VS Code. You get the familiar editor, extensions and keybindings, with AI layered on top. ",[203,1836,1837],{},"Tab"," is Cursor's autocomplete, which in Cursor's words \"suggests code as you type, based on your recent edits, surrounding code, and linter errors\". The agent works in a side panel and shows diffs in the editor. Cursor also ships a CLI, installed with:",[732,1840,1842],{"className":734,"code":1841,"language":736,"meta":613,"style":613},"curl https:\u002F\u002Fcursor.com\u002Finstall -fsS | bash\n",[300,1843,1844],{"__ignoreMap":613},[740,1845,1846,1848,1851,1854,1856],{"class":742,"line":743},[740,1847,753],{"class":752},[740,1849,1850],{"class":760}," https:\u002F\u002Fcursor.com\u002Finstall",[740,1852,1853],{"class":756}," -fsS",[740,1855,765],{"class":764},[740,1857,768],{"class":752},[11,1859,1860,1861,1864],{},"The binary is called ",[300,1862,1863],{},"agent",", and it has a print mode for scripts and CI.",[11,1866,1867,1869],{},[203,1868,1821],{}," started as a terminal tool, and the CLI is still the most complete surface:",[732,1871,1873],{"className":734,"code":1872,"language":736,"meta":613,"style":613},"curl -fsSL https:\u002F\u002Fclaude.ai\u002Finstall.sh | bash\ncd your-project\nclaude\n",[300,1874,1875,1887,1894],{"__ignoreMap":613},[740,1876,1877,1879,1881,1883,1885],{"class":742,"line":743},[740,1878,753],{"class":752},[740,1880,757],{"class":756},[740,1882,761],{"class":760},[740,1884,765],{"class":764},[740,1886,768],{"class":752},[740,1888,1889,1891],{"class":742,"line":614},[740,1890,1223],{"class":756},[740,1892,1893],{"class":760}," your-project\n",[740,1895,1896],{"class":742,"line":615},[740,1897,1231],{"class":752},[11,1899,1900,1901,1903,1904,827],{},"Around that engine Anthropic now ships a VS Code extension (inline diffs, @-mentions, plan review, conversation history), a JetBrains plugin, a desktop app where you run sessions side by side and review diffs visually, and Claude Code on the web at claude.ai\u002Fcode. Anthropic's docs say every surface \"connects to the same underlying Claude Code engine\", so a repo's ",[300,1902,719],{},", settings and MCP servers carry across. Installation details are in ",[18,1905,1906],{"href":825},"how to install Claude Code",[11,1908,1909],{},"The practical gap: Claude Code has no Tab-style autocomplete. If inline completion is a big part of how your team writes code, that matters.",[26,1911,1913],{"id":1912},"models","Models",[34,1915,1916,1926],{},[37,1917,1918],{},[40,1919,1920,1922,1924],{},[43,1921],{},[43,1923,1815],{},[43,1925,1821],{},[59,1927,1928,1939,1950,1965],{},[40,1929,1930,1933,1936],{},[64,1931,1932],{},"Model providers",[64,1934,1935],{},"Anthropic, OpenAI, Google and others",[64,1937,1938],{},"Anthropic only",[40,1940,1941,1944,1947],{},[64,1942,1943],{},"Own model",[64,1945,1946],{},"Composer (currently 2.5)",[64,1948,1949],{},"No",[40,1951,1952,1955,1958],{},[64,1953,1954],{},"Automatic model choice",[64,1956,1957],{},"Auto, with Cost, Balance and Intelligence modes",[64,1959,1960,1961,1964],{},"Aliases such as ",[300,1962,1963],{},"opusplan"," (Opus for planning, Sonnet for execution)",[40,1966,1967,1970,1973],{},[64,1968,1969],{},"Hosting options",[64,1971,1972],{},"Cursor's service",[64,1974,1975],{},"Anthropic API, Amazon Bedrock, Google Cloud, Microsoft Foundry, LLM gateways",[11,1977,1978],{},"Cursor's strength is choice: you can run the same task on Claude, GPT or Gemini and keep whichever works best. Claude Code's strength is depth: the harness is tuned for one model family by the people who train it, and you choose where that model is hosted. For EU companies that last point is often decisive, because it lets you keep inference in an EU region.",[11,1980,1981],{},"One billing detail people miss: when you use Claude inside Cursor, it is billed against your Cursor plan at the model's API rate. Your Claude subscription does not apply there.",[26,1983,1985],{"id":1984},"agent-capabilities","Agent capabilities",[11,1987,1988],{},"Both tools now do the core agent loop well: read the codebase, plan, edit many files, run commands and tests, fix what fails, and commit. The differences are in what surrounds that loop.",[11,1990,1991],{},[203,1992,1815],{},[197,1994,1995,2001,2007],{},[200,1996,1997,2000],{},[203,1998,1999],{},"Cloud agents"," run in isolated cloud VMs with a full development environment. You can start them from the desktop app, cursor.com\u002Fagents, the iOS app, Slack, GitHub or Bitbucket comments, Linear and an API.",[200,2002,2003,2006],{},[203,2004,2005],{},"Bugbot"," does agentic code review on pull requests.",[200,2008,2009,521,2012,342,2015,2018,2019,2022],{},[203,2010,2011],{},"Hooks",[203,2013,2014],{},"rules",[300,2016,2017],{},".cursor\u002Frules",", ",[300,2020,2021],{},"AGENTS.md",") shape agent behaviour.",[11,2024,2025],{},[203,2026,1821],{},[197,2028,2029,2035,2044,2054],{},[200,2030,2031,2034],{},[203,2032,2033],{},"Subagents"," split work across several agents, with a lead agent coordinating.",[200,2036,2037,342,2040,2043],{},[203,2038,2039],{},"Headless mode",[300,2041,2042],{},"claude -p",") pipes into scripts, and there are official GitHub Actions and GitLab CI\u002FCD integrations.",[200,2045,2046,2049,2050,2053],{},[203,2047,2048],{},"Routines"," run on a schedule in the cloud; ",[300,2051,2052],{},"\u002Floop"," repeats a prompt in a session.",[200,2055,2056,2058,2059,2062],{},[203,2057,2011],{}," run shell commands before or after actions, and the ",[203,2060,2061],{},"Agent SDK"," lets you build your own agents on the same tooling.",[732,2064,2066],{"className":734,"code":2065,"language":736,"meta":613,"style":613},"git diff main --name-only | claude -p \"review these changed files for security issues\"\n",[300,2067,2068],{"__ignoreMap":613},[740,2069,2070,2073,2076,2079,2082,2084,2087,2090],{"class":742,"line":743},[740,2071,2072],{"class":752},"git",[740,2074,2075],{"class":760}," diff",[740,2077,2078],{"class":760}," main",[740,2080,2081],{"class":756}," --name-only",[740,2083,765],{"class":764},[740,2085,2086],{"class":752}," claude",[740,2088,2089],{"class":756}," -p",[740,2091,2092],{"class":760}," \"review these changed files for security issues\"\n",[11,2094,2095],{},"Our take: Claude Code is the more programmable of the two. Cursor's cloud agents are the more polished way to hand a ticket to an agent from Slack or Linear and get a branch back.",[26,2097,2099],{"id":2098},"pricing","Pricing",[11,2101,2102],{},"Prices from the vendors' pages, in US dollars per month, as of October 2026.",[34,2104,2105,2117],{},[37,2106,2107],{},[40,2108,2109,2112,2114],{},[43,2110,2111],{},"Tier",[43,2113,1815],{},[43,2115,2116],{},"Claude (includes Claude Code)",[59,2118,2119,2130,2141,2152,2162],{},[40,2120,2121,2124,2127],{},[64,2122,2123],{},"Free",[64,2125,2126],{},"Hobby: limited agent requests",[64,2128,2129],{},"Free plan: Claude Code not included",[40,2131,2132,2135,2138],{},[64,2133,2134],{},"Entry",[64,2136,2137],{},"Pro: $20",[64,2139,2140],{},"Pro: $20 monthly, $17 billed annually",[40,2142,2143,2146,2149],{},[64,2144,2145],{},"Heavy individual",[64,2147,2148],{},"Pro+: $60, Ultra: $200",[64,2150,2151],{},"Max: from $100 (5x or 20x Pro usage)",[40,2153,2154,2156,2159],{},[64,2155,1612],{},[64,2157,2158],{},"Teams Standard: $40\u002Fuser, Teams Premium: $120\u002Fuser",[64,2160,2161],{},"Team Standard: $25\u002Fseat ($20 annual), Team Premium: $125\u002Fseat ($100 annual)",[40,2163,2164,2166,2169],{},[64,2165,1626],{},[64,2167,2168],{},"Custom",[64,2170,2171],{},"$20\u002Fseat billed annually, plus usage at API rates",[11,2173,2174,2175,2177],{},"Both vendors meter usage, not just seats. Cursor says model choice \"affects how quickly your included usage is consumed\", and past the limit you continue at the same API rates. Claude plans have usage windows, and Claude Code can also be billed per token through the API. Our ",[18,2176,586],{"href":585}," goes through those options in detail.",[26,2179,2181],{"id":2180},"mcp-skills-and-rules","MCP, skills and rules",[11,2183,2184],{},"This is where the two have converged most.",[11,2186,2187,2190,2191,2194,2195,2198,2199,2201,2202,2205,2206,521,2209,827],{},[203,2188,2189],{},"MCP."," Both are full MCP clients with local and remote servers. Cursor reads ",[300,2192,2193],{},".cursor\u002Fmcp.json"," per project and ",[300,2196,2197],{},"~\u002F.cursor\u002Fmcp.json"," globally; Claude Code uses ",[300,2200,895],{}," and a shared ",[300,2203,2204],{},".mcp.json",". Setup guides: ",[18,2207,1815],{"href":2208},"\u002Fen\u002Fguides\u002Fmcp\u002Fadd-mcp-server-cursor",[18,2210,1821],{"href":2211},"\u002Fen\u002Fguides\u002Fmcp\u002Fadd-mcp-server-claude-code",[732,2213,2217],{"className":2214,"code":2215,"language":2216,"meta":613,"style":613},"language-json shiki shiki-themes github-dark","{\n  \"mcpServers\": {\n    \"notion\": { \"url\": \"https:\u002F\u002Fmcp.notion.com\u002Fmcp\" }\n  }\n}\n","json",[300,2218,2219,2224,2232,2252,2257],{"__ignoreMap":613},[740,2220,2221],{"class":742,"line":743},[740,2222,2223],{"class":1145},"{\n",[740,2225,2226,2229],{"class":742,"line":614},[740,2227,2228],{"class":756},"  \"mcpServers\"",[740,2230,2231],{"class":1145},": {\n",[740,2233,2234,2237,2240,2243,2246,2249],{"class":742,"line":615},[740,2235,2236],{"class":756},"    \"notion\"",[740,2238,2239],{"class":1145},": { ",[740,2241,2242],{"class":756},"\"url\"",[740,2244,2245],{"class":1145},": ",[740,2247,2248],{"class":760},"\"https:\u002F\u002Fmcp.notion.com\u002Fmcp\"",[740,2250,2251],{"class":1145}," }\n",[740,2253,2254],{"class":742,"line":776},[740,2255,2256],{"class":1145},"  }\n",[740,2258,2259],{"class":742,"line":782},[740,2260,2261],{"class":1145},"}\n",[11,2263,2264,2266,2267,2269,2270,521,2273,2276,2277,521,2280,827],{},[203,2265,909],{}," Both support Agent Skills, folders with a ",[300,2268,913],{},". Cursor loads its own directories and, for compatibility, also ",[300,2271,2272],{},".claude\u002Fskills\u002F",[300,2274,2275],{},"~\u002F.claude\u002Fskills\u002F",". That means one skills folder in the repo can serve both tools. See ",[18,2278,2279],{"href":917},"what skills are",[18,2281,2283],{"href":2282},"\u002Fen\u002Fguides\u002Fskills\u002Fclaude-skills-vs-mcp","skills vs MCP",[11,2285,2286,2289,2290,2292,2293,2295,2296,521,2298,2300,2301,827],{},[203,2287,2288],{},"Instructions."," Claude Code reads ",[300,2291,719],{}," and can also read ",[300,2294,2021],{},". Cursor reads ",[300,2297,2017],{},[300,2299,2021],{},". If you use both tools, keep shared conventions in ",[300,2302,2021],{},[26,2304,2306],{"id":2305},"team-and-enterprise-controls","Team and enterprise controls",[11,2308,2309,2312,2313,2316],{},[203,2310,2311],{},"Cursor Teams"," adds central billing and administration, team-wide privacy mode, SAML\u002FOIDC SSO, usage analytics and a team marketplace for rules, skills and plugins. ",[203,2314,2315],{},"Cursor Enterprise"," adds pooled usage, SCIM, repository, model and MCP access controls, auto-run, browser and network controls, and audit logs.",[11,2318,2319,2321,2322,2325,2326,2329,2330,2333],{},[203,2320,1821],{}," is governed through managed settings: a ",[300,2323,2324],{},"managed-settings.json"," file, an MDM policy or server-managed settings from the claude.ai console. Anthropic's docs say \"nothing you set overrides them\". Admins can lock permissions (",[300,2327,2328],{},"allowManagedPermissionRulesOnly",") and restrict models (",[300,2331,2332],{},"availableModels","). Claude Enterprise adds user and org spend limits and role-based access.",[11,2335,2336],{},"Both are serious about control at the top tier. The difference is scope: each governs its own tool. Neither governs the other, the chat assistant in the browser, or the agents your engineers build with the API.",[26,2338,2340],{"id":2339},"when-to-pick-which","When to pick which",[34,2342,2343,2353],{},[37,2344,2345],{},[40,2346,2347,2350],{},[43,2348,2349],{},"Pick Cursor if",[43,2351,2352],{},"Pick Claude Code if",[59,2354,2355,2363,2371,2379],{},[40,2356,2357,2360],{},[64,2358,2359],{},"Inline autocomplete is central to how people code",[64,2361,2362],{},"Most work is multi-step agent tasks",[40,2364,2365,2368],{},[64,2366,2367],{},"You want to switch between models from several labs",[64,2369,2370],{},"You want Claude, hosted where you choose",[40,2372,2373,2376],{},[64,2374,2375],{},"Your team is happy in a VS Code style editor",[64,2377,2378],{},"Your team uses JetBrains, Vim or mixed editors",[40,2380,2381,2384],{},[64,2382,2383],{},"You want agents launched from Slack, Linear or GitHub with little setup",[64,2385,2386],{},"You want to script the agent into CI, cron and your own tooling",[26,2388,2390],{"id":2389},"using-both","Using both",[11,2392,2393,2394,2396,2397,2018,2399,2401],{},"Running both is common and low friction. Anthropic publishes an install link for the Claude Code extension in Cursor, and the ",[300,2395,821],{}," CLI works in Cursor's integrated terminal. A typical split: Cursor's Tab and quick edits for the moment-to-moment work, Claude Code for larger refactors, test backfills and CI jobs. Keep ",[300,2398,2021],{},[300,2400,2272],{}," and the project MCP config in the repo so both tools read the same conventions.",[11,2403,2404,2405,2408,2409,2413],{},"The cost is governance. Two tools mean two admin consoles, two sets of MCP permissions, two bills and two places where code and prompts leave your network. That is the problem we built Walma AI Hub for: a gateway in your own Azure tenant in an EU region that Claude Code, Cursor and your other AI tools route through, with one model policy, one MCP allowlist, one budget and one log. If that is the question on your desk, the ",[18,2406,2407],{"href":580},"AI Hub page"," has the details, and ",[18,2410,2412],{"href":2411},"\u002Fen\u002Fguides\u002Fmcp\u002Fmcp-gateway","MCP gateways"," explains the pattern.",[1005,2415,2416],{},"html pre.shiki code .svObZ, html code.shiki .svObZ{--shiki-default:#B392F0}html pre.shiki code .sU2Wk, html code.shiki .sU2Wk{--shiki-default:#9ECBFF}html pre.shiki code .sDLfK, html code.shiki .sDLfK{--shiki-default:#79B8FF}html pre.shiki code .snl16, html code.shiki .snl16{--shiki-default:#F97583}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html pre.shiki code .s95oV, html code.shiki .s95oV{--shiki-default:#E1E4E8}",{"title":613,"searchDepth":614,"depth":615,"links":2418},[2419,2420,2421,2422,2423,2424,2425,2426,2427],{"id":1807,"depth":614,"text":1808},{"id":1828,"depth":614,"text":1829},{"id":1912,"depth":614,"text":1913},{"id":1984,"depth":614,"text":1985},{"id":2098,"depth":614,"text":2099},{"id":2180,"depth":614,"text":2181},{"id":2305,"depth":614,"text":2306},{"id":2339,"depth":614,"text":2340},{"id":2389,"depth":614,"text":2390},"2026-10-05","An honest comparison of Cursor and Claude Code as of October 2026: interface, models, agent capabilities, pricing tiers, MCP and skills support, team and enterprise controls, when to pick which, and how to run both side by side.",[2431,2434,2437,2440,2443],{"q":2432,"a":2433},"What is the main difference between Cursor and Claude Code?","Cursor is an AI code editor: a VS Code based IDE with autocomplete, an agent panel and cloud agents built in. Claude Code is Anthropic's coding agent: it started in the terminal and now also runs as a VS Code and JetBrains extension, a desktop app and on the web. Cursor is built around the editor, Claude Code around the agent.",{"q":2435,"a":2436},"Can I use Claude models in Cursor?","Yes. Cursor offers models from Anthropic, OpenAI, Google and others, plus its own Composer model. Claude usage in Cursor is billed against your Cursor plan at the model's API rate, not against a Claude subscription.",{"q":2438,"a":2439},"Can I use GPT or Gemini in Claude Code?","No. Claude Code runs Anthropic's Claude models only. You can reach those models through Anthropic directly or through Amazon Bedrock, Google Cloud, Microsoft Foundry or an LLM gateway, but the models themselves are always Claude.",{"q":2441,"a":2442},"Can I run Claude Code inside Cursor?","Yes. Anthropic's docs list an install link for Cursor, and the Claude Code extension is also on the Open VSX registry. You can also just run the claude CLI in Cursor's integrated terminal.",{"q":2444,"a":2445},"Which is cheaper, Cursor or Claude Code?","Both start at 20 US dollars a month for an individual plan (Cursor Pro, Claude Pro). Heavy users move up to Cursor Pro+ (60) or Ultra (200), or to Claude Max (from 100). For teams, Cursor Teams is 40 per user per month and Claude Team starts at 25 per seat billed monthly. The real cost depends on how much agent work each person runs.",{},"\u002Fguides\u002Fclaude-code\u002Fcursor-vs-claude-code",{"title":1796,"description":2429},"guides\u002Fclaude-code\u002Fcursor-vs-claude-code",null,"yAo9uj_30aqYbxTadAGSbIsxAraaW1zmp-0-P77WpMM",{"id":2453,"title":2454,"author":6,"body":2455,"date":2428,"description":3306,"extension":632,"faq":3307,"meta":3323,"navigation":650,"order":776,"path":3324,"readTime":1525,"seo":3325,"stem":3326,"topic":656,"translationId":2450,"updated":2428,"__hash__":3327},"guides\u002Fguides\u002Fclaude-code\u002Fclaude-code-vs-codex.md","Claude Code vs Codex: an honest comparison for teams (2026)",{"type":8,"value":2456,"toc":3295},[2457,2460,2466,2470,2578,2582,2585,2640,2650,2655,2657,2660,2685,2704,2707,2711,2714,2788,2793,2797,2800,2831,2853,2887,2894,2965,2968,2972,2975,2980,3028,3052,3091,3119,3138,3142,3145,3150,3159,3243,3246,3250,3253,3266,3270,3293],[11,2458,2459],{},"Claude Code and OpenAI Codex have converged. Both started as terminal agents, both now run in the IDE, in a desktop app and in the cloud, and both read project instructions, call MCP servers and load skills. So \"which one is better\" is the wrong question for most teams. The useful questions are which models you want, how much autonomy each tool grants by default, and how either one fits your cloud, your budget and your compliance rules.",[11,2461,2462,2463,827],{},"This guide compares the two as of October 2026, based on the vendors' own documentation. If you are new to Anthropic's tool, start with the ",[18,2464,2465],{"href":1056},"Claude Code overview",[26,2467,2469],{"id":2468},"at-a-glance","At a glance",[34,2471,2472,2483],{},[37,2473,2474],{},[40,2475,2476,2478,2480],{},[43,2477],{},[43,2479,1821],{},[43,2481,2482],{},"Codex",[59,2484,2485,2496,2507,2517,2534,2549,2567],{},[40,2486,2487,2490,2493],{},[64,2488,2489],{},"Vendor",[64,2491,2492],{},"Anthropic",[64,2494,2495],{},"OpenAI",[40,2497,2498,2501,2504],{},[64,2499,2500],{},"Surfaces",[64,2502,2503],{},"Terminal, VS Code, JetBrains, desktop app, web, Slack, GitHub Actions",[64,2505,2506],{},"CLI, IDE extension, desktop app, web\u002Fcloud, GitHub",[40,2508,2509,2511,2514],{},[64,2510,1913],{},[64,2512,2513],{},"Opus, Sonnet, Haiku (Opus 5.5 is the default on paid plans)",[64,2515,2516],{},"GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna",[40,2518,2519,2522,2530],{},[64,2520,2521],{},"Project instructions",[64,2523,2524,2526,2527,2529],{},[300,2525,719],{}," (reads ",[300,2528,2021],{}," too)",[64,2531,2532],{},[300,2533,2021],{},[40,2535,2536,2539,2544],{},[64,2537,2538],{},"Skills folder",[64,2540,2541],{},[300,2542,2543],{},".claude\u002Fskills",[64,2545,2546],{},[300,2547,2548],{},".agents\u002Fskills",[40,2550,2551,2554,2560],{},[64,2552,2553],{},"Local sandbox",[64,2555,2556,2557],{},"Opt-in, ",[300,2558,2559],{},"\u002Fsandbox",[64,2561,2562,2563,2566],{},"On by default (",[300,2564,2565],{},"workspace-write",")",[40,2568,2569,2572,2575],{},[64,2570,2571],{},"Cheapest paid plan",[64,2573,2574],{},"Pro, 20 USD\u002Fmonth",[64,2576,2577],{},"Plus, 20 USD\u002Fmonth (also on Free and Go)",[26,2579,2581],{"id":2580},"workflow-and-surfaces","Workflow and surfaces",[11,2583,2584],{},"Day to day, the two feel alike. You start the agent in a repository, describe a task, and it searches the code, plans, edits files and runs tests.",[732,2586,2588],{"className":734,"code":2587,"language":736,"meta":613,"style":613},"# Claude Code\ncurl -fsSL https:\u002F\u002Fclaude.ai\u002Finstall.sh | bash\nclaude\n\n# Codex\ncurl -fsSL https:\u002F\u002Fchatgpt.com\u002Fcodex\u002Finstall.sh | sh\ncodex\n",[300,2589,2590,2595,2607,2611,2615,2620,2635],{"__ignoreMap":613},[740,2591,2592],{"class":742,"line":743},[740,2593,2594],{"class":746},"# Claude Code\n",[740,2596,2597,2599,2601,2603,2605],{"class":742,"line":614},[740,2598,753],{"class":752},[740,2600,757],{"class":756},[740,2602,761],{"class":760},[740,2604,765],{"class":764},[740,2606,768],{"class":752},[740,2608,2609],{"class":742,"line":615},[740,2610,1231],{"class":752},[740,2612,2613],{"class":742,"line":776},[740,2614,773],{"emptyLinePlaceholder":650},[740,2616,2617],{"class":742,"line":782},[740,2618,2619],{"class":746},"# Codex\n",[740,2621,2623,2625,2627,2630,2632],{"class":742,"line":2622},6,[740,2624,753],{"class":752},[740,2626,757],{"class":756},[740,2628,2629],{"class":760}," https:\u002F\u002Fchatgpt.com\u002Fcodex\u002Finstall.sh",[740,2631,765],{"class":764},[740,2633,2634],{"class":752}," sh\n",[740,2636,2637],{"class":742,"line":651},[740,2638,2639],{"class":752},"codex\n",[11,2641,2642,2643,521,2646,2649],{},"Both have a non-interactive mode for scripts and CI: ",[300,2644,2645],{},"claude -p \"...\"",[300,2647,2648],{},"codex exec \"...\"",". Both run long tasks in the cloud while your laptop is closed, Claude Code on the web and in the Claude mobile app, Codex in its web and cloud surface. Claude Code adds scheduled routines, a Slack integration that turns a bug report into a pull request, and subagents that split a task in parallel. Codex is tightly tied to the ChatGPT account and apps, which matters if your organisation already lives in ChatGPT.",[11,2651,2652,2653,827],{},"Installation details for Claude Code, including Windows, are in ",[18,2654,826],{"href":825},[26,2656,1913],{"id":1912},[11,2658,2659],{},"This is the real difference, and the reason the comparison never settles.",[11,2661,2662,2664,2665,2668,2669,2672,2673,2676,2677,2680,2681,2684],{},[203,2663,1821],{}," runs Anthropic's models. The ",[300,2666,2667],{},"opus"," alias currently maps to Opus 5.5 on Anthropic's API, ",[300,2670,2671],{},"sonnet"," to Sonnet 5.5, and ",[300,2674,2675],{},"haiku"," covers quick, cheap work. Switch with ",[300,2678,2679],{},"\u002Fmodel"," in a session or ",[300,2682,2683],{},"claude --model sonnet"," at startup. Note that the mapping differs by provider: on Amazon Bedrock and Microsoft Foundry the same aliases point to older versions, so check what your cloud actually serves.",[11,2686,2687,2689,2690,2692,2693,2696,2697,2700,2701,827],{},[203,2688,2482],{}," runs OpenAI's models. OpenAI's docs recommend GPT-6 Astra as the most capable, GPT-6.1 Sol as near-Astra at lower cost, and GPT-6 Luna for focused, efficient tasks. GPT-5.5 retires on 14 October 2026. Switch with ",[300,2691,2679],{}," or ",[300,2694,2695],{},"codex -m gpt-6.1-sol",", or set ",[300,2698,2699],{},"model"," in ",[300,2702,2703],{},"~\u002F.codex\u002Fconfig.toml",[11,2705,2706],{},"In practice developers develop strong preferences per task type: one model plans a refactor better, the other writes the tests faster, and the ranking shifts with every release. We are not going to publish a winner here, because any benchmark in a guide like this is out of date within a quarter. Run both on a week of your own tickets and look at the diffs.",[26,2708,2710],{"id":2709},"pricing-and-plans","Pricing and plans",[11,2712,2713],{},"Neither tool is sold on its own. Both come with a chat subscription or with per-token API billing.",[34,2715,2716,2727],{},[37,2717,2718],{},[40,2719,2720,2723,2725],{},[43,2721,2722],{},"Plan",[43,2724,1821],{},[43,2726,2482],{},[59,2728,2729,2739,2749,2759,2769,2779],{},[40,2730,2731,2733,2736],{},[64,2732,2123],{},[64,2734,2735],{},"Not included",[64,2737,2738],{},"Included (ChatGPT Free)",[40,2740,2741,2743,2746],{},[64,2742,2134],{},[64,2744,2745],{},"Pro, 20 USD\u002Fmonth (17 billed annually)",[64,2747,2748],{},"Go, 8 USD\u002Fmonth; Plus, 20 USD\u002Fmonth",[40,2750,2751,2753,2756],{},[64,2752,2145],{},[64,2754,2755],{},"Max, from 100 USD\u002Fmonth",[64,2757,2758],{},"Pro, from 100 USD\u002Fmonth",[40,2760,2761,2763,2766],{},[64,2762,1612],{},[64,2764,2765],{},"Team standard seat 20 USD (annual) or 25 USD (monthly); premium seat 100\u002F125 USD",[64,2767,2768],{},"Business, 20 USD\u002Fuser (annual) or 25 USD (monthly)",[40,2770,2771,2773,2776],{},[64,2772,1626],{},[64,2774,2775],{},"20 USD\u002Fseat\u002Fmonth plus usage at API rates",[64,2777,2778],{},"Custom pricing",[40,2780,2781,2784,2786],{},[64,2782,2783],{},"API key",[64,2785,1645],{},[64,2787,1645],{},[11,2789,2790,2791,827],{},"Usage limits are where the real cost hides. Both vendors meter subscription usage in rolling windows, and both say consumption depends heavily on model and task size. A developer who runs the agent for hours a day will outgrow the 20 dollar tier on either side. For Claude Code numbers and cost controls, see ",[18,2792,605],{"href":585},[26,2794,2796],{"id":2795},"sandboxing-and-approvals","Sandboxing and approvals",[11,2798,2799],{},"Here the defaults differ, and it is worth understanding before you roll either out.",[11,2801,2802,2805,2806,2018,2809,2811,2812,2815,2816,2818,2819,2822,2823,2826,2827,2830],{},[203,2803,2804],{},"Codex sandboxes by default."," It has three sandbox modes: ",[300,2807,2808],{},"read-only",[300,2810,2565],{}," (the default for local work: edit the workspace, run local commands inside it) and ",[300,2813,2814],{},"danger-full-access",". In ",[300,2817,2565],{},", outbound network access is off unless you opt in. Approval policies decide when the agent stops to ask: ",[300,2820,2821],{},"on-request"," asks when it needs to go beyond the sandbox, ",[300,2824,2825],{},"never"," does not stop. Enforcement is native: Seatbelt on macOS, ",[300,2828,2829],{},"bubblewrap"," on Linux and WSL2.",[732,2832,2834],{"className":734,"code":2833,"language":736,"meta":613,"style":613},"codex --sandbox workspace-write --ask-for-approval on-request\n",[300,2835,2836],{"__ignoreMap":613},[740,2837,2838,2841,2844,2847,2850],{"class":742,"line":743},[740,2839,2840],{"class":752},"codex",[740,2842,2843],{"class":756}," --sandbox",[740,2845,2846],{"class":760}," workspace-write",[740,2848,2849],{"class":756}," --ask-for-approval",[740,2851,2852],{"class":760}," on-request\n",[11,2854,2855,2858,2859,2018,2862,2865,2866,2869,2870,521,2873,2876,2877,2879,2880,2882,2883,2886],{},[203,2856,2857],{},"Claude Code asks by default and sandboxes on request."," Its default mode prompts before edits and shell commands. Other modes are ",[300,2860,2861],{},"acceptEdits",[300,2863,2864],{},"plan"," (read-only exploration), ",[300,2867,2868],{},"auto"," (a classifier reviews actions instead of you), ",[300,2871,2872],{},"dontAsk",[300,2874,2875],{},"bypassPermissions",". The OS-level sandbox for shell commands uses the same Seatbelt and ",[300,2878,2829],{}," mechanisms but is off by default: run ",[300,2881,2559],{}," or set ",[300,2884,2885],{},"sandbox.enabled"," in settings. It covers shell commands only; file tools, MCP servers and hooks run outside it.",[11,2888,2889,2890,2893],{},"For a team, Claude Code's strength is central policy. Administrators deploy managed settings that users cannot override, including deny rules and ",[300,2891,2892],{},"disableBypassPermissionsMode",":",[732,2895,2897],{"className":2214,"code":2896,"language":2216,"meta":613,"style":613},"{\n  \"permissions\": {\n    \"deny\": [\"Read(.\u002F.env)\", \"Bash(curl:*)\"],\n    \"disableBypassPermissionsMode\": \"disable\"\n  },\n  \"sandbox\": { \"enabled\": true }\n}\n",[300,2898,2899,2903,2910,2929,2939,2944,2961],{"__ignoreMap":613},[740,2900,2901],{"class":742,"line":743},[740,2902,2223],{"class":1145},[740,2904,2905,2908],{"class":742,"line":614},[740,2906,2907],{"class":756},"  \"permissions\"",[740,2909,2231],{"class":1145},[740,2911,2912,2915,2918,2921,2923,2926],{"class":742,"line":615},[740,2913,2914],{"class":756},"    \"deny\"",[740,2916,2917],{"class":1145},": [",[740,2919,2920],{"class":760},"\"Read(.\u002F.env)\"",[740,2922,2018],{"class":1145},[740,2924,2925],{"class":760},"\"Bash(curl:*)\"",[740,2927,2928],{"class":1145},"],\n",[740,2930,2931,2934,2936],{"class":742,"line":776},[740,2932,2933],{"class":756},"    \"disableBypassPermissionsMode\"",[740,2935,2245],{"class":1145},[740,2937,2938],{"class":760},"\"disable\"\n",[740,2940,2941],{"class":742,"line":782},[740,2942,2943],{"class":1145},"  },\n",[740,2945,2946,2949,2951,2954,2956,2959],{"class":742,"line":2622},[740,2947,2948],{"class":756},"  \"sandbox\"",[740,2950,2239],{"class":1145},[740,2952,2953],{"class":756},"\"enabled\"",[740,2955,2245],{"class":1145},[740,2957,2958],{"class":756},"true",[740,2960,2251],{"class":1145},[740,2962,2963],{"class":742,"line":651},[740,2964,2261],{"class":1145},[11,2966,2967],{},"Our view: Codex has the safer out-of-the-box default for an individual developer; Claude Code gives an organisation finer, enforceable rules. Either way, set the policy centrally rather than trusting each laptop.",[26,2969,2971],{"id":2970},"mcp-skills-and-instruction-files","MCP, skills and instruction files",[11,2973,2974],{},"The ecosystems have largely standardised, which is good news if you run both.",[11,2976,2977,2979],{},[203,2978,2189],{}," Both are MCP clients for stdio and streamable HTTP servers, with OAuth for remote ones. The same server, added to each:",[732,2981,2983],{"className":734,"code":2982,"language":736,"meta":613,"style":613},"claude mcp add context7 -- npx -y @upstash\u002Fcontext7-mcp\ncodex mcp add context7 -- npx -y @upstash\u002Fcontext7-mcp\n",[300,2984,2985,3010],{"__ignoreMap":613},[740,2986,2987,2989,2992,2995,2998,3001,3004,3007],{"class":742,"line":743},[740,2988,821],{"class":752},[740,2990,2991],{"class":760}," mcp",[740,2993,2994],{"class":760}," add",[740,2996,2997],{"class":760}," context7",[740,2999,3000],{"class":756}," --",[740,3002,3003],{"class":760}," npx",[740,3005,3006],{"class":756}," -y",[740,3008,3009],{"class":760}," @upstash\u002Fcontext7-mcp\n",[740,3011,3012,3014,3016,3018,3020,3022,3024,3026],{"class":742,"line":614},[740,3013,2840],{"class":752},[740,3015,2991],{"class":760},[740,3017,2994],{"class":760},[740,3019,2997],{"class":760},[740,3021,3000],{"class":756},[740,3023,3003],{"class":760},[740,3025,3006],{"class":756},[740,3027,3009],{"class":760},[11,3029,3030,3031,3033,3034,3037,3038,3040,3041,3044,3045,3048,3049,827],{},"Claude Code stores project servers in ",[300,3032,2204],{},"; Codex uses ",[300,3035,3036],{},"[mcp_servers.\u003Cname>]"," blocks in ",[300,3039,2703],{}," or a project ",[300,3042,3043],{},".codex\u002Fconfig.toml",", and ",[300,3046,3047],{},"codex mcp login \u003Cname>"," for OAuth. More in ",[18,3050,3051],{"href":2211},"Add an MCP server to Claude Code",[11,3053,3054,3056,3057,3059,3060,3063,3064,3067,3068,521,3070,3073,3074,521,3076,3079,3080,2692,3083,3086,3087,3090],{},[203,3055,909],{}," Both follow the open Agent Skills standard: a folder with a ",[300,3058,913],{}," holding a ",[300,3061,3062],{},"name",", a ",[300,3065,3066],{},"description"," and instructions. Claude Code reads ",[300,3069,2543],{},[300,3071,3072],{},"~\u002F.claude\u002Fskills","; Codex reads ",[300,3075,2548],{},[300,3077,3078],{},"~\u002F.agents\u002Fskills",". Claude Code adds its own frontmatter fields (such as ",[300,3081,3082],{},"context: fork",[300,3084,3085],{},"disable-model-invocation",") that Codex will ignore, so keep shared skills to the standard fields. See ",[18,3088,3089],{"href":2282},"Claude skills vs MCP"," for when each fits.",[11,3092,3093,3096,3097,3099,3100,3103,3104,3106,3107,3109,3110,3112,3113,3115,3116,3118],{},[203,3094,3095],{},"AGENTS.md vs CLAUDE.md."," Codex reads ",[300,3098,2021],{}," from ",[300,3101,3102],{},"~\u002F.codex",", then from the git root down to the current directory, capped at 32 KiB by default. Claude Code reads ",[300,3105,719],{},", and since recent versions also reads ",[300,3108,2021],{}," when there is no ",[300,3111,719],{},". The simplest setup for a mixed team is one ",[300,3114,2021],{}," as the source of truth, plus a one-line ",[300,3117,719],{}," that imports it:",[732,3120,3122],{"className":734,"code":3121,"language":736,"meta":613,"style":613},"echo \"@AGENTS.md\" > CLAUDE.md\n",[300,3123,3124],{"__ignoreMap":613},[740,3125,3126,3129,3132,3135],{"class":742,"line":743},[740,3127,3128],{"class":756},"echo",[740,3130,3131],{"class":760}," \"@AGENTS.md\"",[740,3133,3134],{"class":764}," >",[740,3136,3137],{"class":760}," CLAUDE.md\n",[26,3139,3141],{"id":3140},"enterprise-data-and-residency","Enterprise, data and residency",[11,3143,3144],{},"At the enterprise tier OpenAI lists SCIM, role-based access, audit logs through its Compliance API and no training on business data by default. On the Anthropic side, Team and Enterprise come with central admin and the managed settings shown above. The bigger difference for European companies is where inference runs.",[11,3146,3147,3149],{},[203,3148,1821],{}," can be pointed at Amazon Bedrock, Google Cloud or Microsoft Foundry instead of Anthropic's API, so prompts and code are processed in the cloud region you pick. Model availability varies by provider and region.",[11,3151,3152,3154,3155,3158],{},[203,3153,2482],{}," can use a custom model provider. OpenAI's docs show an Azure OpenAI example in ",[300,3156,3157],{},"config.toml",", which sends requests to your own Azure resource:",[732,3160,3164],{"className":3161,"code":3162,"language":3163,"meta":613,"style":613},"language-toml shiki shiki-themes github-dark","# ~\u002F.codex\u002Fconfig.toml\nmodel_provider = \"azure\"\n\n[model_providers.azure]\nname = \"Azure\"\nbase_url = \"https:\u002F\u002FYOUR_PROJECT_NAME.openai.azure.com\u002Fopenai\"\nenv_key = \"AZURE_OPENAI_API_KEY\"\nquery_params = { api-version = \"2025-04-01-preview\" }\nwire_api = \"responses\"\n","toml",[300,3165,3166,3171,3179,3183,3199,3207,3215,3223,3234],{"__ignoreMap":613},[740,3167,3168],{"class":742,"line":743},[740,3169,3170],{"class":746},"# ~\u002F.codex\u002Fconfig.toml\n",[740,3172,3173,3176],{"class":742,"line":614},[740,3174,3175],{"class":1145},"model_provider = ",[740,3177,3178],{"class":760},"\"azure\"\n",[740,3180,3181],{"class":742,"line":615},[740,3182,773],{"emptyLinePlaceholder":650},[740,3184,3185,3188,3191,3193,3196],{"class":742,"line":776},[740,3186,3187],{"class":1145},"[",[740,3189,3190],{"class":752},"model_providers",[740,3192,827],{"class":1145},[740,3194,3195],{"class":752},"azure",[740,3197,3198],{"class":1145},"]\n",[740,3200,3201,3204],{"class":742,"line":782},[740,3202,3203],{"class":1145},"name = ",[740,3205,3206],{"class":760},"\"Azure\"\n",[740,3208,3209,3212],{"class":742,"line":2622},[740,3210,3211],{"class":1145},"base_url = ",[740,3213,3214],{"class":760},"\"https:\u002F\u002FYOUR_PROJECT_NAME.openai.azure.com\u002Fopenai\"\n",[740,3216,3217,3220],{"class":742,"line":651},[740,3218,3219],{"class":1145},"env_key = ",[740,3221,3222],{"class":760},"\"AZURE_OPENAI_API_KEY\"\n",[740,3224,3226,3229,3232],{"class":742,"line":3225},8,[740,3227,3228],{"class":1145},"query_params = { api-version = ",[740,3230,3231],{"class":760},"\"2025-04-01-preview\"",[740,3233,2251],{"class":1145},[740,3235,3237,3240],{"class":742,"line":3236},9,[740,3238,3239],{"class":1145},"wire_api = ",[740,3241,3242],{"class":760},"\"responses\"\n",[11,3244,3245],{},"Check which deployment type each model has in your region: the newest OpenAI models do not always launch in EU data zones.",[26,3247,3249],{"id":3248},"the-multi-model-argument","The multi-model argument",[11,3251,3252],{},"Most teams we talk to do not pick one. Some developers prefer Claude for long refactors, others prefer GPT for quick fixes, and the preference flips with each model release. Standardising on one vendor saves an invoice but locks you out of whichever model is ahead next quarter.",[11,3254,3255,3256,3258,3259,3261,3262,3265],{},"Running both is easy on the developer side, since the MCP servers, skills and ",[300,3257,2021],{}," carry over. The hard part is governance: two sets of keys, two policy models, two audit trails, two places code leaves the building. That is the job of a gateway. ",[18,3260,581],{"href":580}," runs in your own Azure tenant in an EU region and serves Claude, GPT and other models to Claude Code, Codex and Cursor through one endpoint, with per-team budgets, model policy, approved MCP servers and one log. Our ",[18,3263,3264],{"href":2411},"MCP gateway guide"," explains the pattern in more depth.",[26,3267,3269],{"id":3268},"which-should-you-choose","Which should you choose?",[197,3271,3272,3278,3284],{},[200,3273,3274,3277],{},[203,3275,3276],{},"Choose Claude Code"," if you want fine-grained, centrally enforced permissions, routines and Slack-driven workflows, or if your developers prefer Claude models.",[200,3279,3280,3283],{},[203,3281,3282],{},"Choose Codex"," if your company already standardised on ChatGPT, you want a sandbox that is on from the first run, or your developers prefer GPT models.",[200,3285,3286,3289,3290,3292],{},[203,3287,3288],{},"Choose both"," if you have more than a handful of developers. Share one ",[300,3291,2021],{},", keep skills to the open standard, and put policy and logging in a gateway rather than on each machine.",[1005,3294,1494],{},{"title":613,"searchDepth":614,"depth":615,"links":3296},[3297,3298,3299,3300,3301,3302,3303,3304,3305],{"id":2468,"depth":614,"text":2469},{"id":2580,"depth":614,"text":2581},{"id":1912,"depth":614,"text":1913},{"id":2709,"depth":614,"text":2710},{"id":2795,"depth":614,"text":2796},{"id":2970,"depth":614,"text":2971},{"id":3140,"depth":614,"text":3141},{"id":3248,"depth":614,"text":3249},{"id":3268,"depth":614,"text":3269},"Claude Code vs OpenAI Codex compared on what matters to a team: surfaces and workflow, models, pricing, sandboxing and approvals, MCP, skills and AGENTS.md vs CLAUDE.md, and enterprise data residency. Plus why many teams end up running both.",[3308,3311,3314,3317,3320],{"q":3309,"a":3310},"What is the difference between Claude Code and Codex?","Both are agentic coding tools that read a repository, edit files and run commands. Claude Code is Anthropic's agent and runs Claude models (Opus, Sonnet, Haiku). Codex is OpenAI's agent and runs GPT models (GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna). The workflows are now very similar; the bigger differences are the models, the default safety posture and how each plugs into your cloud.",{"q":3312,"a":3313},"Is Codex cheaper than Claude Code?","The entry points are the same: both are included in a 20 US dollar per month individual plan, and both offer business seats from 20 dollars per user per month billed annually. Codex is also available on ChatGPT Free and Go. What you actually pay depends on usage limits, model choice and whether you bill per token through the API.",{"q":3315,"a":3316},"Can Claude Code read AGENTS.md?","Yes. Recent versions of Claude Code read AGENTS.md when the repository has no CLAUDE.md, and a setting lets it read both. You can also import AGENTS.md from a CLAUDE.md, so one instruction file can serve both agents.",{"q":3318,"a":3319},"Do Claude Code and Codex both support MCP and skills?","Yes. Both connect to MCP servers over stdio and HTTP, and both support skills in the open Agent Skills format (a folder with a SKILL.md file). Claude Code looks in .claude\u002Fskills, Codex in .agents\u002Fskills.",{"q":3321,"a":3322},"Can we run Claude Code and Codex with EU data residency?","Claude Code can run against Amazon Bedrock, Google Cloud or Microsoft Foundry instead of Anthropic's API, and Codex can point at an Azure OpenAI resource through a custom model provider. In both cases inference follows the cloud region you choose, subject to which models are deployed there. A gateway in your own EU tenant can serve both.",{},"\u002Fguides\u002Fclaude-code\u002Fclaude-code-vs-codex",{"title":2454,"description":3306},"guides\u002Fclaude-code\u002Fclaude-code-vs-codex","piJFFu4V2HArpes1pLF-OIZf_hlQdXCrW40RDiOPVho",{"id":3329,"title":3330,"author":6,"body":3331,"date":2428,"description":4261,"extension":632,"faq":4262,"meta":4278,"navigation":650,"order":782,"path":4279,"readTime":1525,"seo":4280,"stem":4281,"topic":656,"translationId":2450,"updated":2428,"__hash__":4282},"guides\u002Fguides\u002Fclaude-code\u002Fhow-to-use-claude-code.md","How to use Claude Code: a hands-on tutorial for daily work",{"type":8,"value":3332,"toc":4247},[3333,3342,3346,3349,3363,3366,3384,3388,3395,3486,3492,3501,3552,3559,3563,3566,3604,3607,3611,3617,3629,3679,3692,3696,3702,3783,3797,3801,3804,3837,3840,3843,3859,3912,3915,3933,4043,4056,4060,4066,4091,4099,4127,4131,4134,4186,4189,4193,4230,4234,4244],[11,3334,3335,3336,3338,3339,3341],{},"This tutorial starts where installation ends. If you have not installed the agent yet, follow ",[18,3337,826],{"href":825},", and if you want the big picture of what it is and what it costs, read the ",[18,3340,2465],{"href":1056},". What follows is how to work with it day to day: the controls you touch every session, the files that make it better over time, and a full worked example. Everything here reflects Anthropic's Claude Code docs as of October 2026.",[26,3343,3345],{"id":3344},"your-first-session","Your first session",[11,3347,3348],{},"Start the agent from the root of a repository:",[732,3350,3351],{"className":734,"code":1216,"language":736,"meta":613,"style":613},[300,3352,3353,3359],{"__ignoreMap":613},[740,3354,3355,3357],{"class":742,"line":743},[740,3356,1223],{"class":756},[740,3358,1226],{"class":760},[740,3360,3361],{"class":742,"line":614},[740,3362,1231],{"class":752},[11,3364,3365],{},"Start with questions rather than code: \"give me an overview of this codebase\", \"how is authentication handled?\". You learn the repo and see how Claude navigates it at the same time.",[11,3367,3368,3369,342,3372,3375,3376,3379,3380,3383],{},"Reference files with ",[300,3370,3371],{},"@",[300,3373,3374],{},"@src\u002Futils\u002Fauth.js",") instead of describing where code lives, and paste screenshots of errors or mockups straight into the prompt. To pick a session back up later, ",[300,3377,3378],{},"claude --continue"," resumes the most recent conversation in that directory and ",[300,3381,3382],{},"claude --resume"," shows a list.",[26,3385,3387],{"id":3386},"permission-modes","Permission modes",[11,3389,3390,3391,3394],{},"A permission mode decides what Claude may do without asking. Press ",[203,3392,3393],{},"Shift+Tab"," to cycle through them during a session; the status bar shows the active one.",[34,3396,3397,3410],{},[37,3398,3399],{},[40,3400,3401,3404,3407],{},[43,3402,3403],{},"Mode",[43,3405,3406],{},"Runs without asking",[43,3408,3409],{},"Use it for",[59,3411,3412,3426,3438,3450,3462,3474],{},[40,3413,3414,3420,3423],{},[64,3415,3416,3417,2566],{},"Manual (",[300,3418,3419],{},"default",[64,3421,3422],{},"Reads only",[64,3424,3425],{},"Sensitive work, unfamiliar code",[40,3427,3428,3432,3435],{},[64,3429,3430],{},[300,3431,2861],{},[64,3433,3434],{},"Reads, file edits, common filesystem commands",[64,3436,3437],{},"Iterating on code you are reviewing",[40,3439,3440,3444,3447],{},[64,3441,3442],{},[300,3443,2864],{},[64,3445,3446],{},"Reads; no edits until you approve a plan",[64,3448,3449],{},"Exploring before changing anything",[40,3451,3452,3456,3459],{},[64,3453,3454],{},[300,3455,2868],{},[64,3457,3458],{},"Most actions, reviewed by a classifier model",[64,3460,3461],{},"Long tasks with fewer interruptions",[40,3463,3464,3468,3471],{},[64,3465,3466],{},[300,3467,2872],{},[64,3469,3470],{},"Only pre-approved tools, everything else denied",[64,3472,3473],{},"Locked-down CI and scripts",[40,3475,3476,3480,3483],{},[64,3477,3478],{},[300,3479,2875],{},[64,3481,3482],{},"Everything",[64,3484,3485],{},"Isolated containers and VMs only",[11,3487,3488,3489,827],{},"In current versions, auto mode is the starting mode for interactive terminal and VS Code sessions, provided you use a supported model and your organization has not turned it off. To approve every step yourself, start with ",[300,3490,3491],{},"claude --permission-mode manual",[11,3493,3494,3495,3497,3498,2893],{},"Modes set the baseline. On top of them, ",[300,3496,1330],{}," manages allow, ask and deny rules. Deny rules apply in every mode, so this is the place to block things like pushing to ",[300,3499,3500],{},"main",[732,3502,3504],{"className":2214,"code":3503,"language":2216,"meta":613,"style":613},"{\n  \"permissions\": {\n    \"allow\": [\"Bash(npm run test *)\", \"Bash(npm run lint)\"],\n    \"deny\": [\"Bash(git push origin main)\"]\n  }\n}\n",[300,3505,3506,3510,3516,3533,3544,3548],{"__ignoreMap":613},[740,3507,3508],{"class":742,"line":743},[740,3509,2223],{"class":1145},[740,3511,3512,3514],{"class":742,"line":614},[740,3513,2907],{"class":756},[740,3515,2231],{"class":1145},[740,3517,3518,3521,3523,3526,3528,3531],{"class":742,"line":615},[740,3519,3520],{"class":756},"    \"allow\"",[740,3522,2917],{"class":1145},[740,3524,3525],{"class":760},"\"Bash(npm run test *)\"",[740,3527,2018],{"class":1145},[740,3529,3530],{"class":760},"\"Bash(npm run lint)\"",[740,3532,2928],{"class":1145},[740,3534,3535,3537,3539,3542],{"class":742,"line":776},[740,3536,2914],{"class":756},[740,3538,2917],{"class":1145},[740,3540,3541],{"class":760},"\"Bash(git push origin main)\"",[740,3543,3198],{"class":1145},[740,3545,3546],{"class":742,"line":782},[740,3547,2256],{"class":1145},[740,3549,3550],{"class":742,"line":2622},[740,3551,2261],{"class":1145},[11,3553,3554,3555,3558],{},"That block goes in ",[300,3556,3557],{},".claude\u002Fsettings.json"," and is shared with the team through git.",[26,3560,3562],{"id":3561},"plan-mode-explore-plan-then-code","Plan mode: explore, plan, then code",[11,3564,3565],{},"For anything that touches several files, separate thinking from typing. Anthropic's recommended loop has four steps:",[537,3567,3568,3582,3592,3598],{},[200,3569,3570,3573,3574,3577,3578,3581],{},[203,3571,3572],{},"Explore"," in plan mode (Shift+Tab until ",[300,3575,3576],{},"plan mode on",", or ",[300,3579,3580],{},"\u002Fplan"," in front of one prompt). Claude reads and answers without editing.",[200,3583,3584,3587,3588,3591],{},[203,3585,3586],{},"Plan."," Ask what needs to change and for a step-by-step plan. Press ",[203,3589,3590],{},"Ctrl+G"," to open the plan in your editor and fix it directly.",[200,3593,3594,3597],{},[203,3595,3596],{},"Implement."," Approve the plan; the session leaves plan mode and Claude starts editing, checking its work against the plan.",[200,3599,3600,3603],{},[203,3601,3602],{},"Commit."," Ask for a descriptive commit and a pull request.",[11,3605,3606],{},"Skip the plan when you could describe the diff in one sentence.",[26,3608,3610],{"id":3609},"claudemd-and-init","CLAUDE.md and \u002Finit",[11,3612,3613,3614,3616],{},"Every session starts with an empty context window. ",[300,3615,719],{}," is how knowledge survives between sessions: Claude loads it at the start of every conversation.",[11,3618,3619,3620,3622,3623,3625,3626,3628],{},"Run ",[300,3621,879],{}," once per repository. It analyzes the codebase and drafts a ",[300,3624,719],{}," with build commands, test instructions and conventions it finds. If one already exists, it suggests improvements instead of overwriting. Then edit by hand. Keep commands Claude cannot guess, style rules that differ from defaults, PR conventions and gotchas. Leave out anything Claude can learn by reading the code, and keep each file under about 200 lines. A bloated ",[300,3627,719],{}," gets followed less, not more. Where the files live:",[34,3630,3631,3641],{},[37,3632,3633],{},[40,3634,3635,3638],{},[43,3636,3637],{},"File",[43,3639,3640],{},"Scope",[59,3642,3643,3656,3669],{},[40,3644,3645,3653],{},[64,3646,3647,2692,3650],{},[300,3648,3649],{},".\u002FCLAUDE.md",[300,3651,3652],{},".\u002F.claude\u002FCLAUDE.md",[64,3654,3655],{},"The project, committed for the team",[40,3657,3658,3663],{},[64,3659,3660],{},[300,3661,3662],{},".\u002FCLAUDE.local.md",[64,3664,3665,3666,2566],{},"Your personal notes for this project (add to ",[300,3667,3668],{},".gitignore",[40,3670,3671,3676],{},[64,3672,3673],{},[300,3674,3675],{},"~\u002F.claude\u002FCLAUDE.md",[64,3677,3678],{},"You, across all projects",[11,3680,3681,3684,3685,3688,3689,3691],{},[300,3682,3683],{},"\u002Fmemory"," opens these files for editing, and ",[300,3686,3687],{},"@path\u002Fto\u002Ffile"," inside a ",[300,3690,719],{}," imports another file.",[26,3693,3695],{"id":3694},"the-slash-commands-you-will-actually-use","The slash commands you will actually use",[11,3697,3698,3699,3701],{},"Type ",[300,3700,1155],{}," to see everything. These are the ones that come up daily:",[34,3703,3704,3713],{},[37,3705,3706],{},[40,3707,3708,3710],{},[43,3709,1289],{},[43,3711,3712],{},"What it does",[59,3714,3715,3724,3734,3744,3754,3764,3773],{},[40,3716,3717,3721],{},[64,3718,3719],{},[300,3720,1340],{},[64,3722,3723],{},"New conversation, empty context",[40,3725,3726,3731],{},[64,3727,3728],{},[300,3729,3730],{},"\u002Fcompact [focus]",[64,3732,3733],{},"Summarize the conversation to free space",[40,3735,3736,3741],{},[64,3737,3738],{},[300,3739,3740],{},"\u002Fcontext",[64,3742,3743],{},"Show what is filling the context window",[40,3745,3746,3751],{},[64,3747,3748],{},[300,3749,3750],{},"\u002Frewind",[64,3752,3753],{},"Restore conversation and\u002For code to an earlier checkpoint (also Esc Esc)",[40,3755,3756,3761],{},[64,3757,3758],{},[300,3759,3760],{},"\u002Fplan [task]",[64,3762,3763],{},"Enter plan mode, optionally with a task",[40,3765,3766,3770],{},[64,3767,3768],{},[300,3769,899],{},[64,3771,3772],{},"Manage MCP servers and their OAuth logins",[40,3774,3775,3780],{},[64,3776,3777],{},[300,3778,3779],{},"\u002Fcode-review",[64,3781,3782],{},"Review the current diff for bugs in a fresh subagent",[11,3784,3785,3786,3789,3790,3793,3794,827],{},"Your own commands are now skills (more on that below): ",[300,3787,3788],{},".claude\u002Fcommands\u002Fdeploy.md"," still works and creates ",[300,3791,3792],{},"\u002Fdeploy",", exactly like ",[300,3795,3796],{},".claude\u002Fskills\u002Fdeploy\u002FSKILL.md",[26,3798,3800],{"id":3799},"managing-context","Managing context",[11,3802,3803],{},"The context window is the resource that matters most. Every file read and every command output lands in it, and quality drops as it fills. The habits that keep sessions sharp:",[197,3805,3806,3814,3823,3831],{},[200,3807,3808,3813],{},[203,3809,3810,3812],{},[300,3811,1340],{}," between unrelated tasks."," Mixing three topics in one session is the most common failure.",[200,3815,3816,3819,3820,3822],{},[203,3817,3818],{},"Correct early."," Esc stops Claude mid-action with context intact. After two failed corrections, ",[300,3821,1340],{}," and write a better prompt.",[200,3824,3825,3830],{},[203,3826,3827],{},[300,3828,3829],{},"\u002Fcompact Focus on the API changes"," when you want to keep going but trim history.",[200,3832,3833,3836],{},[203,3834,3835],{},"Rewind instead of arguing."," Checkpoints are taken on every prompt, so you can let Claude try something risky and roll back. They only track edits made through Claude's own file tools, not Bash, so git remains your real safety net.",[26,3838,2033],{"id":3839},"subagents",[11,3841,3842],{},"A subagent runs in its own context window and reports back a summary, so exploration does not fill your main session. Built-in ones include Explore (read-only search) and Plan. The simplest use is to ask: \"use a subagent to investigate how our auth system handles token refresh\".",[11,3844,3845,3846,3849,3850,3853,3854,521,3856,3858],{},"Custom subagents are Markdown files in ",[300,3847,3848],{},".claude\u002Fagents\u002F"," (project) or ",[300,3851,3852],{},"~\u002F.claude\u002Fagents\u002F"," (personal). Only ",[300,3855,3062],{},[300,3857,3066],{}," are required in the frontmatter; the body is the agent's instructions:",[732,3860,3864],{"className":3861,"code":3862,"language":3863,"meta":613,"style":613},"language-yaml shiki shiki-themes github-dark","---\nname: security-reviewer\ndescription: Reviews code for security vulnerabilities\ntools: Read, Grep, Glob, Bash\nmodel: opus\n---\n","yaml",[300,3865,3866,3871,3881,3890,3899,3908],{"__ignoreMap":613},[740,3867,3868],{"class":742,"line":743},[740,3869,3870],{"class":752},"---\n",[740,3872,3873,3876,3878],{"class":742,"line":614},[740,3874,3062],{"class":3875},"s4JwU",[740,3877,2245],{"class":1145},[740,3879,3880],{"class":760},"security-reviewer\n",[740,3882,3883,3885,3887],{"class":742,"line":615},[740,3884,3066],{"class":3875},[740,3886,2245],{"class":1145},[740,3888,3889],{"class":760},"Reviews code for security vulnerabilities\n",[740,3891,3892,3894,3896],{"class":742,"line":776},[740,3893,283],{"class":3875},[740,3895,2245],{"class":1145},[740,3897,3898],{"class":760},"Read, Grep, Glob, Bash\n",[740,3900,3901,3903,3905],{"class":742,"line":782},[740,3902,2699],{"class":3875},[740,3904,2245],{"class":1145},[740,3906,3907],{"class":760},"opus\n",[740,3909,3910],{"class":742,"line":2622},[740,3911,3870],{"class":752},[26,3913,2011],{"id":3914},"hooks",[11,3916,3917,3919,3920,2018,3923,2018,3926,2692,3929,3932],{},[300,3918,719],{}," is advice. A hook is a guarantee: a shell command Claude Code runs at a fixed point in its lifecycle, such as ",[300,3921,3922],{},"PreToolUse",[300,3924,3925],{},"PostToolUse",[300,3927,3928],{},"Notification",[300,3930,3931],{},"Stop",". This example from Anthropic's docs runs Prettier on every file Claude edits:",[732,3934,3936],{"className":2214,"code":3935,"language":2216,"meta":613,"style":613},"{\n  \"hooks\": {\n    \"PostToolUse\": [\n      {\n        \"matcher\": \"Edit|Write\",\n        \"hooks\": [\n          {\n            \"type\": \"command\",\n            \"command\": \"jq -r '.tool_input.file_path' | xargs npx prettier --write\"\n          }\n        ]\n      }\n    ]\n  }\n}\n",[300,3937,3938,3942,3949,3957,3962,3975,3982,3987,3999,4009,4015,4021,4027,4033,4038],{"__ignoreMap":613},[740,3939,3940],{"class":742,"line":743},[740,3941,2223],{"class":1145},[740,3943,3944,3947],{"class":742,"line":614},[740,3945,3946],{"class":756},"  \"hooks\"",[740,3948,2231],{"class":1145},[740,3950,3951,3954],{"class":742,"line":615},[740,3952,3953],{"class":756},"    \"PostToolUse\"",[740,3955,3956],{"class":1145},": [\n",[740,3958,3959],{"class":742,"line":776},[740,3960,3961],{"class":1145},"      {\n",[740,3963,3964,3967,3969,3972],{"class":742,"line":782},[740,3965,3966],{"class":756},"        \"matcher\"",[740,3968,2245],{"class":1145},[740,3970,3971],{"class":760},"\"Edit|Write\"",[740,3973,3974],{"class":1145},",\n",[740,3976,3977,3980],{"class":742,"line":2622},[740,3978,3979],{"class":756},"        \"hooks\"",[740,3981,3956],{"class":1145},[740,3983,3984],{"class":742,"line":651},[740,3985,3986],{"class":1145},"          {\n",[740,3988,3989,3992,3994,3997],{"class":742,"line":3225},[740,3990,3991],{"class":756},"            \"type\"",[740,3993,2245],{"class":1145},[740,3995,3996],{"class":760},"\"command\"",[740,3998,3974],{"class":1145},[740,4000,4001,4004,4006],{"class":742,"line":3236},[740,4002,4003],{"class":756},"            \"command\"",[740,4005,2245],{"class":1145},[740,4007,4008],{"class":760},"\"jq -r '.tool_input.file_path' | xargs npx prettier --write\"\n",[740,4010,4012],{"class":742,"line":4011},10,[740,4013,4014],{"class":1145},"          }\n",[740,4016,4018],{"class":742,"line":4017},11,[740,4019,4020],{"class":1145},"        ]\n",[740,4022,4024],{"class":742,"line":4023},12,[740,4025,4026],{"class":1145},"      }\n",[740,4028,4030],{"class":742,"line":4029},13,[740,4031,4032],{"class":1145},"    ]\n",[740,4034,4036],{"class":742,"line":4035},14,[740,4037,2256],{"class":1145},[740,4039,4041],{"class":742,"line":4040},15,[740,4042,2261],{"class":1145},[11,4044,4045,4046,4048,4049,4052,4053,4055],{},"Put it in ",[300,4047,3557],{}," and check it with ",[300,4050,4051],{},"\u002Fhooks",". If Claude keeps breaking a rule in ",[300,4054,719],{},", turn it into a hook.",[26,4057,4059],{"id":4058},"mcp-and-skills-in-brief","MCP and skills in brief",[11,4061,4062,4065],{},[203,4063,4064],{},"MCP servers"," connect Claude Code to external systems such as issue trackers, databases and design tools:",[732,4067,4069],{"className":734,"code":4068,"language":736,"meta":613,"style":613},"claude mcp add --transport http notion https:\u002F\u002Fmcp.notion.com\u002Fmcp\n",[300,4070,4071],{"__ignoreMap":613},[740,4072,4073,4075,4077,4079,4082,4085,4088],{"class":742,"line":743},[740,4074,821],{"class":752},[740,4076,2991],{"class":760},[740,4078,2994],{"class":760},[740,4080,4081],{"class":756}," --transport",[740,4083,4084],{"class":760}," http",[740,4086,4087],{"class":760}," notion",[740,4089,4090],{"class":760}," https:\u002F\u002Fmcp.notion.com\u002Fmcp\n",[11,4092,818,4093,4095,4096,827],{},[300,4094,899],{}," inside a session to log in. The full walkthrough is in ",[18,4097,4098],{"href":2211},"adding an MCP server to Claude Code",[11,4100,4101,4104,4105,2700,4107,4110,4111,4113,4114,4116,4117,4120,4121,4124,4125,827],{},[203,4102,4103],{},"Skills"," are folders with a ",[300,4106,913],{},[300,4108,4109],{},".claude\u002Fskills\u002F\u003Cname>\u002F"," (or ",[300,4112,2275],{}," for personal ones). Claude loads one automatically when its ",[300,4115,3066],{}," matches the task, or you run it as ",[300,4118,4119],{},"\u002Fname",". Add ",[300,4122,4123],{},"disable-model-invocation: true"," for workflows with side effects, like deploys, that should only run when you type the command. See ",[18,4126,918],{"href":917},[26,4128,4130],{"id":4129},"worked-example-from-bug-report-to-pull-request","Worked example: from bug report to pull request",[11,4132,4133],{},"A realistic flow for a bug report that says login fails after a session timeout:",[537,4135,4136,4145,4151,4157,4163,4172],{},[200,4137,4138,4141,4142,4144],{},[203,4139,4140],{},"Start clean."," ",[300,4143,1340],{},", or a new session. In plan mode, prompt: \"users report that login fails after session timeout. check the auth flow in src\u002Fauth\u002F, especially token refresh. explain the likely cause\".",[200,4146,4147,4150],{},[203,4148,4149],{},"Reproduce first."," \"Write a failing test that reproduces the issue. Run it and show me the output.\" A failing test is the check Claude will iterate against.",[200,4152,4153,4156],{},[203,4154,4155],{},"Plan and approve."," Ask for a minimal fix plan, adjust it with Ctrl+G if needed, approve.",[200,4158,4159,4162],{},[203,4160,4161],{},"Fix and verify."," \"Implement the fix, run the auth tests and the linter, fix any failures. Address the root cause, do not suppress the error.\" Ask Claude to show the test output rather than just claim success.",[200,4164,4165,4168,4169,4171],{},[203,4166,4167],{},"Review."," Run ",[300,4170,3779],{}," so a fresh subagent looks at the diff without the reasoning that produced it.",[200,4173,4174,4177,4178,4181,4182,4185],{},[203,4175,4176],{},"Ship."," \"Commit with a descriptive message and open a PR.\" With ",[300,4179,4180],{},"gh"," installed and authenticated, Claude creates the PR and links the session to it; ",[300,4183,4184],{},"claude --from-pr \u003Cnumber>"," finds that session again later.",[11,4187,4188],{},"The key move is step 2: give Claude a check it can run, and it closes the loop itself instead of waiting for you to spot mistakes.",[26,4190,4192],{"id":4191},"tips-that-actually-help","Tips that actually help",[197,4194,4195,4201,4207,4217],{},[200,4196,4197,4200],{},[203,4198,4199],{},"Be specific."," Name files, the scenario and what \"done\" looks like. \"Add tests for foo.py\" is weaker than \"write a test for foo.py covering the logged-out case, avoid mocks\".",[200,4202,4203,4206],{},[203,4204,4205],{},"Point to existing patterns"," in the repo rather than describing them.",[200,4208,4209,4212,4213,4216],{},[203,4210,4211],{},"Run parallel work in worktrees"," with ",[300,4214,4215],{},"claude --worktree feature-auth"," so edits do not collide.",[200,4218,4219,4212,4222,4225,4226,4229],{},[203,4220,4221],{},"Name sessions",[300,4223,4224],{},"\u002Frename"," so ",[300,4227,4228],{},"\u002Fresume"," stays useful.",[26,4231,4233],{"id":4232},"rolling-it-out-to-a-team","Rolling it out to a team",[11,4235,4236,4237,4240,4241,4243],{},"For a team, the open questions are where traffic goes, which MCP servers are allowed and who can see what was done. Claude Code can run against an LLM gateway instead of Anthropic directly, which puts model access, MCP policy and logging in one place. Walma AI Hub is that gateway, running in the customer's own Azure tenant in an EU region; see ",[18,4238,4239],{"href":580},"the AI Hub"," or the ",[18,4242,3264],{"href":2411}," for how the controls fit together.",[1005,4245,4246],{},"html pre.shiki code .sDLfK, html code.shiki .sDLfK{--shiki-default:#79B8FF}html pre.shiki code .sU2Wk, html code.shiki .sU2Wk{--shiki-default:#9ECBFF}html pre.shiki code .svObZ, html code.shiki .svObZ{--shiki-default:#B392F0}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html pre.shiki code .s95oV, html code.shiki .s95oV{--shiki-default:#E1E4E8}html pre.shiki code .s4JwU, html code.shiki .s4JwU{--shiki-default:#85E89D}",{"title":613,"searchDepth":614,"depth":615,"links":4248},[4249,4250,4251,4252,4253,4254,4255,4256,4257,4258,4259,4260],{"id":3344,"depth":614,"text":3345},{"id":3386,"depth":614,"text":3387},{"id":3561,"depth":614,"text":3562},{"id":3609,"depth":614,"text":3610},{"id":3694,"depth":614,"text":3695},{"id":3799,"depth":614,"text":3800},{"id":3839,"depth":614,"text":2033},{"id":3914,"depth":614,"text":2011},{"id":4058,"depth":614,"text":4059},{"id":4129,"depth":614,"text":4130},{"id":4191,"depth":614,"text":4192},{"id":4232,"depth":614,"text":4233},"A practical Claude Code tutorial for after installation: your first session, permission modes, plan mode, CLAUDE.md and \u002Finit, the slash commands that matter, context management, subagents, hooks, MCP and skills, plus a worked example from bug to tested pull request.",[4263,4266,4269,4272,4275],{"q":4264,"a":4265},"How do I start using Claude Code?","Open a terminal in your project folder and run claude. Ask for an overview of the codebase first, then run \u002Finit to generate a CLAUDE.md with your build and test commands. From there, describe tasks in plain language and let Claude read, edit and run commands, reviewing as it goes.",{"q":4267,"a":4268},"What are the permission modes in Claude Code?","Manual (config value default) asks before edits and commands. acceptEdits auto-approves file edits and common filesystem commands. plan lets Claude read and propose a plan without editing. auto lets a classifier model review actions instead of you. dontAsk only runs pre-approved tools, and bypassPermissions skips checks and is meant for isolated containers. Press Shift+Tab to cycle modes during a session.",{"q":4270,"a":4271},"How do I use plan mode in Claude Code?","Press Shift+Tab until the status bar shows plan mode on, prefix a single prompt with \u002Fplan, or start with claude --permission-mode plan. Claude explores and writes a plan without editing your source. Press Ctrl+G to edit the plan in your text editor, then approve it to let Claude start implementing.",{"q":4273,"a":4274},"What is the difference between \u002Fclear and \u002Fcompact?","\u002Fclear starts a new conversation with an empty context, which is what you want between unrelated tasks. \u002Fcompact keeps the same conversation but summarizes it to free space, and accepts focus instructions such as \u002Fcompact Focus on the API changes.",{"q":4276,"a":4277},"Do custom slash commands still exist in Claude Code?","Yes, but they have been merged into skills. A file at .claude\u002Fcommands\u002Fdeploy.md and a skill at .claude\u002Fskills\u002Fdeploy\u002FSKILL.md both create \u002Fdeploy. Existing command files keep working; skills add supporting files and frontmatter that controls whether Claude can invoke them on its own.",{},"\u002Fguides\u002Fclaude-code\u002Fhow-to-use-claude-code",{"title":3330,"description":4261},"guides\u002Fclaude-code\u002Fhow-to-use-claude-code","gMQmoqbyVHoKoOd1_cg0XiiqSCuHEC29LnPtD7AS0EE",{"id":4284,"title":4285,"author":6,"body":4286,"date":630,"description":4750,"extension":632,"faq":4751,"meta":4767,"navigation":650,"order":2622,"path":4768,"readTime":1525,"seo":4769,"stem":4770,"topic":656,"translationId":4771,"updated":630,"__hash__":4772},"guides\u002Fguides\u002Fclaude-code\u002Fclaude-api-key.md","Claude API key: how to get one, use it and keep it safe",{"type":8,"value":4287,"toc":4741},[4288,4291,4300,4304,4315,4324,4328,4370,4376,4380,4383,4440,4447,4457,4461,4468,4484,4487,4564,4581,4591,4600,4604,4617,4620,4646,4649,4653,4656,4659,4698,4718,4720,4738],[11,4289,4290],{},"A Claude API key is what lets your code, scripts and agents call Claude directly, outside the Claude.ai apps. Getting one takes a minute. Choosing the right kind of key, and keeping it from ending up in a Git repository or a Slack thread, takes a bit more thought.",[11,4292,4293,4294,4299],{},"This guide covers how to create a key, which type to choose, how to use it, and how a team should manage keys. Facts are from Anthropic's documentation as of October 2026; the Console changes often, so check ",[18,4295,4298],{"href":4296,"rel":4297},"https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fget-api-key",[22],"Get your Claude API key"," if a screen looks different.",[26,4301,4303],{"id":4302},"before-you-start-the-api-is-not-your-claude-subscription","Before you start: the API is not your Claude subscription",[11,4305,4306,4307,4310,4311,4314],{},"A common surprise: a paid Claude.ai plan (Pro, Max, Team or Enterprise) does ",[203,4308,4309],{},"not"," include API access. The apps and the API are separate products with separate billing. API usage is billed per token in the ",[203,4312,4313],{},"Claude Console",", usually through prepaid credits, and new accounts get a small amount of free credit for testing. Some organisations are invoiced monthly instead through an agreement with Anthropic's sales team.",[11,4316,4317,4318,4323],{},"So you need a Console account at ",[18,4319,4322],{"href":4320,"rel":4321},"https:\u002F\u002Fplatform.claude.com",[22],"platform.claude.com"," even if you already pay for Claude.",[26,4325,4327],{"id":4326},"how-to-create-a-claude-api-key","How to create a Claude API key",[537,4329,4330,4336,4342,4360],{},[200,4331,4332,4335],{},[203,4333,4334],{},"Sign in"," to the Claude Console at platform.claude.com, or create an account.",[200,4337,4338,4339,827],{},"Go to ",[203,4340,4341],{},"Settings → API keys",[200,4343,4344,4345,4348,4349,4352,4353,4356,4357,827],{},"Click ",[203,4346,4347],{},"Create key",". Name it after what will use it (\"ci-release-notes\", not \"test\"), choose an ",[203,4350,4351],{},"expiration",", and set ",[203,4354,4355],{},"Linked account"," to yourself or a service account. Optionally scope it to a ",[203,4358,4359],{},"workspace",[200,4361,4362,4365,4366,4369],{},[203,4363,4364],{},"Copy the key and store it"," in a secrets manager or password manager. It starts with ",[300,4367,4368],{},"sk-ant-",", and the Console shows it only once. If you lose it, you create a new one.",[11,4371,4372,4373,4375],{},"If the ",[203,4374,4347],{}," button is greyed out, your role does not allow it. Ask an organisation admin to change your role or to create a service account key for you.",[26,4377,4379],{"id":4378},"which-key-type-to-choose","Which key type to choose",[11,4381,4382],{},"When you create a key, you choose what it acts as. This matters more than it looks.",[34,4384,4385,4398],{},[37,4386,4387],{},[40,4388,4389,4392,4395],{},[43,4390,4391],{},"Key type",[43,4393,4394],{},"Acts as",[43,4396,4397],{},"Stops working when",[59,4399,4400,4413,4426],{},[40,4401,4402,4407,4410],{},[64,4403,4404],{},[203,4405,4406],{},"Personal key",[64,4408,4409],{},"You, with your roles and permissions",[64,4411,4412],{},"You leave the organisation or lose access to its workspace",[40,4414,4415,4420,4423],{},[64,4416,4417],{},[203,4418,4419],{},"Service account key",[64,4421,4422],{},"A service account (a non-human identity an admin creates)",[64,4424,4425],{},"The service account is archived or removed from the workspace",[40,4427,4428,4434,4437],{},[64,4429,4430,4433],{},[203,4431,4432],{},"Workspace key"," (legacy)",[64,4435,4436],{},"Nobody: it belongs to the workspace",[64,4438,4439],{},"It expires, is disabled or deleted, or the workspace is archived",[11,4441,4442,4443,4446],{},"The rule of thumb from Anthropic's docs is simple: ",[203,4444,4445],{},"a personal key for your own development, a service account key for anything shared"," (CI pipelines, production services, agents). A personal key used by a production service breaks the day that person leaves, and every request in the logs looks like it came from them. Workspace keys still work but are legacy: they outlive the people who created them, which is exactly what you do not want from a credential nobody owns.",[11,4448,4449,4450,4453,4454,4456],{},"For production workloads on AWS, Google Cloud, Azure, GitHub Actions or Kubernetes, Anthropic also supports ",[203,4451,4452],{},"Workload Identity Federation",": the workload exchanges its cloud identity token for a short-lived Claude access token, so there is no static ",[300,4455,4368],{}," string to leak or rotate at all. If your platform supports it, it is the better option for anything that runs unattended.",[26,4458,4460],{"id":4459},"how-to-use-your-api-key","How to use your API key",[11,4462,4463,4464,4467],{},"The SDKs read the ",[300,4465,4466],{},"ANTHROPIC_API_KEY"," environment variable automatically:",[732,4469,4471],{"className":734,"code":4470,"language":736,"meta":613,"style":613},"export ANTHROPIC_API_KEY=\"sk-ant-api03-...\"\n",[300,4472,4473],{"__ignoreMap":613},[740,4474,4475,4477,4479,4481],{"class":742,"line":743},[740,4476,1247],{"class":764},[740,4478,1250],{"class":1145},[740,4480,1253],{"class":764},[740,4482,4483],{"class":760},"\"sk-ant-api03-...\"\n",[11,4485,4486],{},"A first request with curl:",[732,4488,4490],{"className":734,"code":4489,"language":736,"meta":613,"style":613},"curl https:\u002F\u002Fapi.anthropic.com\u002Fv1\u002Fmessages \\\n  -H \"Authorization: Bearer $ANTHROPIC_API_KEY\" \\\n  -H \"anthropic-version: 2023-06-01\" \\\n  -H \"content-type: application\u002Fjson\" \\\n  -d '{\n    \"model\": \"claude-sonnet-5-5\",\n    \"max_tokens\": 1024,\n    \"messages\": [{\"role\": \"user\", \"content\": \"Hello, Claude\"}]\n  }'\n",[300,4491,4492,4502,4518,4527,4536,4544,4549,4554,4559],{"__ignoreMap":613},[740,4493,4494,4496,4499],{"class":742,"line":743},[740,4495,753],{"class":752},[740,4497,4498],{"class":760}," https:\u002F\u002Fapi.anthropic.com\u002Fv1\u002Fmessages",[740,4500,4501],{"class":756}," \\\n",[740,4503,4504,4507,4510,4513,4516],{"class":742,"line":614},[740,4505,4506],{"class":756},"  -H",[740,4508,4509],{"class":760}," \"Authorization: Bearer ",[740,4511,4512],{"class":1145},"$ANTHROPIC_API_KEY",[740,4514,4515],{"class":760},"\"",[740,4517,4501],{"class":756},[740,4519,4520,4522,4525],{"class":742,"line":615},[740,4521,4506],{"class":756},[740,4523,4524],{"class":760}," \"anthropic-version: 2023-06-01\"",[740,4526,4501],{"class":756},[740,4528,4529,4531,4534],{"class":742,"line":776},[740,4530,4506],{"class":756},[740,4532,4533],{"class":760}," \"content-type: application\u002Fjson\"",[740,4535,4501],{"class":756},[740,4537,4538,4541],{"class":742,"line":782},[740,4539,4540],{"class":756},"  -d",[740,4542,4543],{"class":760}," '{\n",[740,4545,4546],{"class":742,"line":2622},[740,4547,4548],{"class":760},"    \"model\": \"claude-sonnet-5-5\",\n",[740,4550,4551],{"class":742,"line":651},[740,4552,4553],{"class":760},"    \"max_tokens\": 1024,\n",[740,4555,4556],{"class":742,"line":3225},[740,4557,4558],{"class":760},"    \"messages\": [{\"role\": \"user\", \"content\": \"Hello, Claude\"}]\n",[740,4560,4561],{"class":742,"line":3236},[740,4562,4563],{"class":760},"  }'\n",[11,4565,4566,4569,4570,4573,4574,521,4577,4580],{},[300,4567,4568],{},"Authorization: Bearer"," is the current header; the older ",[300,4571,4572],{},"x-api-key"," header still works. Both ",[300,4575,4576],{},"anthropic-version",[300,4578,4579],{},"content-type"," are required.",[11,4582,4583,4586,4587,4590],{},[203,4584,4585],{},"Workspaces."," If the key is scoped to one workspace, requests run there automatically. If it can reach several workspaces, every request must also send ",[300,4588,4589],{},"anthropic-workspace-id"," with the workspace's ID (found under Settings → Workspaces), otherwise the API returns a 400 error. For most integrations, scoping each key to one workspace is simpler and safer.",[11,4592,4593,4596,4597,4599],{},[203,4594,4595],{},"Claude Code."," Claude Code can run on an API key instead of a subscription login, which is how it is usually run in CI and automation. Cost then follows API pricing; see ",[18,4598,605],{"href":585}," for how that compares with a subscription.",[26,4601,4603],{"id":4602},"expiration-rotation-and-what-to-do-after-a-leak","Expiration, rotation and what to do after a leak",[11,4605,4606,4607,4610,4611,4613,4614,827],{},"When you create a key you choose an expiration: 3 hours, 1 day, 7 days, 30 days, a custom duration, or ",[203,4608,4609],{},"Never"," (for keys kept in a secrets manager that you rotate yourself). An organisation can set a maximum expiration policy, in which case ",[203,4612,4609],{}," disappears. Expiration is fixed at creation and cannot be changed later. Anthropic emails the creator before longer-lived keys expire (7 days before for keys of 14 days or more, 1 day before for keys of 7 days or more); shorter keys expire silently with a ",[300,4615,4616],{},"401 authentication_error",[11,4618,4619],{},"If a key leaks, act in this order:",[537,4621,4622,4628,4634,4640],{},[200,4623,4624,4627],{},[203,4625,4626],{},"Disable or delete it"," on the API keys page. Disable is reversible, delete is permanent.",[200,4629,4630,4633],{},[203,4631,4632],{},"Create and deploy a replacement",", ideally of the right type this time.",[200,4635,4636,4639],{},[203,4637,4638],{},"Check the Usage page"," for spend you do not recognise.",[200,4641,4642,4645],{},[203,4643,4644],{},"Find out how it leaked"," (a commit, a log line, a screenshot, a frontend bundle) and close that path.",[11,4647,4648],{},"Short expirations for experiments and service accounts for production make step 1 rarely necessary.",[26,4650,4652],{"id":4651},"how-a-team-should-manage-claude-api-keys","How a team should manage Claude API keys",[11,4654,4655],{},"One developer with one key is easy. Twenty developers, three CI pipelines and a handful of agents is where it goes wrong: keys shared in chat, nobody knows which key is used where, and the first sign of a problem is the invoice.",[11,4657,4658],{},"What works:",[197,4660,4661,4671,4677,4682,4688],{},[200,4662,4663,4666,4667,4670],{},[203,4664,4665],{},"One workspace per environment or use case"," (development, CI, production, each agent). Workspaces separate usage, and you can set ",[203,4668,4669],{},"spend and rate limits per workspace"," so one runaway script cannot consume the whole organisation's budget.",[200,4672,4673,4676],{},[203,4674,4675],{},"Service accounts for everything automated",", one per workload, with keys scoped to that workload's workspace.",[200,4678,4679],{},[203,4680,4681],{},"Expirations on everything that is not in a secrets manager.",[200,4683,4684,4687],{},[203,4685,4686],{},"Know your tier's spend cap."," Organisations are placed on usage tiers with a monthly cap (Start $500, Build $1,000, Scale $200,000 at the time of writing). When the cap is reached, the API returns 429 errors until the next month or a tier increase, which can stop production. Set your own lower limit on the Billing page if you want a guardrail, not a cliff.",[200,4689,4690,4693,4694,4697],{},[203,4691,4692],{},"Audit regularly."," The Admin API lists every key with its status and ",[300,4695,4696],{},"expires_at",", so a script can flag keys that never expire or belong to people who left.",[11,4699,4700,4701,2692,4704,4707,4708,4711,4712,4714,4715,827],{},"The harder problem is that every key is a direct line to Anthropic, with whatever data the request carries, in whatever region Anthropic routes it. The Claude API's inference geography options are ",[300,4702,4703],{},"global",[300,4705,4706],{},"us","; there is no EU-only option on the first-party API. Companies that need EU processing usually run Claude through a cloud provider's EU region (Amazon Bedrock or Google Cloud regional endpoints), and put a ",[203,4709,4710],{},"gateway"," in front so developers and agents never hold provider keys at all. The gateway holds the credentials, enforces budgets and model policy per team, and logs every call. That is what ",[18,4713,581],{"href":580}," does, inside the customer's own Azure tenant in an EU region, for Claude, OpenAI models and approved MCP servers. For the general pattern, see ",[18,4716,4717],{"href":2411},"What is an MCP gateway?",[26,4719,591],{"id":590},[197,4721,4722,4728,4733],{},[200,4723,4724,4727],{},[18,4725,4726],{"href":1675},"Claude API pricing",": what your key will cost per model, and how to keep it down.",[200,4729,4730,4732],{},[18,4731,319],{"href":329},": Anthropic's hosted agent harness, which runs on the same key.",[200,4734,4735,4737],{},[18,4736,826],{"href":825},": including running it on an API key.",[1005,4739,4740],{},"html pre.shiki code .snl16, html code.shiki .snl16{--shiki-default:#F97583}html pre.shiki code .s95oV, html code.shiki .s95oV{--shiki-default:#E1E4E8}html pre.shiki code .sU2Wk, html code.shiki .sU2Wk{--shiki-default:#9ECBFF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html pre.shiki code .svObZ, html code.shiki .svObZ{--shiki-default:#B392F0}html pre.shiki code .sDLfK, html code.shiki .sDLfK{--shiki-default:#79B8FF}",{"title":613,"searchDepth":614,"depth":615,"links":4742},[4743,4744,4745,4746,4747,4748,4749],{"id":4302,"depth":614,"text":4303},{"id":4326,"depth":614,"text":4327},{"id":4378,"depth":614,"text":4379},{"id":4459,"depth":614,"text":4460},{"id":4602,"depth":614,"text":4603},{"id":4651,"depth":614,"text":4652},{"id":590,"depth":614,"text":591},"How to create a Claude API key in the Claude Console, which key type to choose (personal, service account or legacy workspace key), how to use it with curl and the SDKs, how expiration and workspaces work, and how a team should manage keys so one leak does not become a bill.",[4752,4755,4758,4761,4764],{"q":4753,"a":4754},"How do I get a Claude API key?","Sign in to the Claude Console at platform.claude.com, open Settings → API keys and click Create key. Give it a name, choose an expiration and a linked account (yourself or a service account), optionally scope it to a workspace, and copy the key. The Console shows the full key, which starts with sk-ant-, only once.",{"q":4756,"a":4757},"Is the Claude API key free?","Creating a key is free, but usage is not. API usage is billed per token through the Console, normally with prepaid credits, and new accounts receive a small amount of free credit for testing. A Claude Pro, Max, Team or Enterprise subscription does not include API access; the two are billed separately.",{"q":4759,"a":4760},"Where do I put my Claude API key?","Set it as the ANTHROPIC_API_KEY environment variable, which the official SDKs read automatically, or load it from a secrets manager. On direct HTTP requests send it as Authorization: Bearer \u003Ckey> (or the legacy x-api-key header). Never put it in source code, a frontend bundle or a shared document.",{"q":4762,"a":4763},"What should I do if my Claude API key leaks?","Disable or delete it immediately on the API keys page in the Console, create a replacement, deploy the new key, and check the Usage page for spend you do not recognise. Disable is reversible; delete is permanent. Setting an expiration and a workspace spend limit in advance limits how much damage a leaked key can do.",{"q":4765,"a":4766},"Can I see my Claude API key again after creating it?","No. The Console shows the full key only once, at creation. If you lose it, create a new key and delete the old one. The Admin API can list keys and their status, but only returns a partially redacted hint, never the secret.",{},"\u002Fguides\u002Fclaude-code\u002Fclaude-api-key",{"title":4285,"description":4750},"guides\u002Fclaude-code\u002Fclaude-api-key","claude-api-key","8fxLDt23olGAiY-R0khH1bynYfQ8QewS9iobdIVhlik",{"id":4,"title":5,"author":6,"body":4774,"date":630,"description":631,"extension":632,"faq":5200,"meta":5206,"navigation":650,"order":651,"path":652,"readTime":653,"seo":5207,"stem":655,"topic":656,"translationId":657,"updated":658,"__hash__":659},{"type":8,"value":4775,"toc":5185},[4776,4778,4783,4785,4787,4901,4903,4905,4919,4921,4923,4925,4927,4957,4959,4961,4963,4965,4985,4987,4993,4995,5009,5011,5015,5019,5069,5071,5075,5077,5079,5109,5113,5115,5117,5131,5133,5135,5137,5163,5169,5171],[11,4777,13],{},[11,4779,16,4780,24],{},[18,4781,23],{"href":20,"rel":4782},[22],[26,4784,29],{"id":28},[11,4786,32],{},[34,4788,4789,4803],{},[37,4790,4791],{},[40,4792,4793,4795,4797,4799,4801],{},[43,4794,45],{},[43,4796,48],{},[43,4798,51],{},[43,4800,54],{},[43,4802,57],{},[59,4804,4805,4817,4829,4841,4853,4865,4877,4889],{},[40,4806,4807,4809,4811,4813,4815],{},[64,4808,66],{},[64,4810,69],{},[64,4812,72],{},[64,4814,75],{},[64,4816,78],{},[40,4818,4819,4821,4823,4825,4827],{},[64,4820,83],{},[64,4822,86],{},[64,4824,89],{},[64,4826,92],{},[64,4828,95],{},[40,4830,4831,4833,4835,4837,4839],{},[64,4832,100],{},[64,4834,103],{},[64,4836,106],{},[64,4838,109],{},[64,4840,112],{},[40,4842,4843,4845,4847,4849,4851],{},[64,4844,117],{},[64,4846,120],{},[64,4848,69],{},[64,4850,92],{},[64,4852,127],{},[40,4854,4855,4857,4859,4861,4863],{},[64,4856,132],{},[64,4858,120],{},[64,4860,69],{},[64,4862,92],{},[64,4864,141],{},[40,4866,4867,4869,4871,4873,4875],{},[64,4868,146],{},[64,4870,149],{},[64,4872,152],{},[64,4874,155],{},[64,4876,158],{},[40,4878,4879,4881,4883,4885,4887],{},[64,4880,163],{},[64,4882,166],{},[64,4884,109],{},[64,4886,171],{},[64,4888,174],{},[40,4890,4891,4893,4895,4897,4899],{},[64,4892,179],{},[64,4894,182],{},[64,4896,103],{},[64,4898,166],{},[64,4900,189],{},[11,4902,192],{},[11,4904,195],{},[197,4906,4907,4911,4915],{},[200,4908,4909,206],{},[203,4910,205],{},[200,4912,4913,212],{},[203,4914,211],{},[200,4916,4917,218],{},[203,4918,217],{},[11,4920,221],{},[26,4922,225],{"id":224},[227,4924,230],{"id":229},[11,4926,233],{},[34,4928,4929,4937],{},[37,4930,4931],{},[40,4932,4933,4935],{},[43,4934,242],{},[43,4936,245],{},[59,4938,4939,4945,4951],{},[40,4940,4941,4943],{},[64,4942,252],{},[64,4944,255],{},[40,4946,4947,4949],{},[64,4948,260],{},[64,4950,263],{},[40,4952,4953,4955],{},[64,4954,54],{},[64,4956,270],{},[11,4958,273],{},[227,4960,277],{"id":276},[11,4962,280],{},[227,4964,284],{"id":283},[197,4966,4967,4971,4977,4981],{},[200,4968,4969,292],{},[203,4970,291],{},[200,4972,4973,298,4975,303],{},[203,4974,297],{},[300,4976,302],{},[200,4978,4979,309],{},[203,4980,308],{},[200,4982,4983,315],{},[203,4984,314],{},[227,4986,319],{"id":318},[11,4988,322,4989,326,4991,330],{},[203,4990,325],{},[18,4992,319],{"href":329},[227,4994,334],{"id":333},[197,4996,4997,5005],{},[200,4998,4999,342,5001,346,5003,350],{},[203,5000,341],{},[300,5002,345],{},[203,5004,349],{},[200,5006,5007,356],{},[203,5008,355],{},[26,5010,360],{"id":359},[11,5012,5013,366],{},[203,5014,365],{},[11,5016,5017,372],{},[203,5018,371],{},[34,5020,5021,5031],{},[37,5022,5023],{},[40,5024,5025,5027,5029],{},[43,5026,381],{},[43,5028,384],{},[43,5030,387],{},[59,5032,5033,5041,5049,5057],{},[40,5034,5035,5037,5039],{},[64,5036,394],{},[64,5038,397],{},[64,5040,400],{},[40,5042,5043,5045,5047],{},[64,5044,405],{},[64,5046,408],{},[64,5048,411],{},[40,5050,5051,5053,5055],{},[64,5052,51],{},[64,5054,418],{},[64,5056,421],{},[40,5058,5059,5063,5065],{},[64,5060,5061],{},[203,5062,428],{},[64,5064],{},[64,5066,5067],{},[203,5068,435],{},[11,5070,438],{},[11,5072,441,5073],{},[203,5074,444],{},[26,5076,448],{"id":447},[11,5078,451],{},[34,5080,5081,5089],{},[37,5082,5083],{},[40,5084,5085,5087],{},[43,5086,460],{},[43,5088,463],{},[59,5090,5091,5097,5103],{},[40,5092,5093,5095],{},[64,5094,470],{},[64,5096,473],{},[40,5098,5099,5101],{},[64,5100,478],{},[64,5102,481],{},[40,5104,5105,5107],{},[64,5106,486],{},[64,5108,489],{},[11,5110,492,5111,496],{},[203,5112,495],{},[26,5114,500],{"id":499},[11,5116,503],{},[197,5118,5119,5125],{},[200,5120,5121,511,5123,515],{},[203,5122,510],{},[203,5124,514],{},[200,5126,5127,521,5129,525],{},[203,5128,520],{},[203,5130,524],{},[11,5132,528],{},[26,5134,532],{"id":531},[11,5136,535],{},[537,5138,5139,5143,5147,5151,5155,5159],{},[200,5140,5141,544],{},[203,5142,543],{},[200,5144,5145,550],{},[203,5146,549],{},[200,5148,5149,556],{},[203,5150,555],{},[200,5152,5153,562],{},[203,5154,561],{},[200,5156,5157,568],{},[203,5158,567],{},[200,5160,5161,574],{},[203,5162,573],{},[11,5164,577,5165,582,5167,587],{},[18,5166,581],{"href":580},[18,5168,586],{"href":585},[26,5170,591],{"id":590},[197,5172,5173,5177,5181],{},[200,5174,5175,600],{},[18,5176,599],{"href":598},[200,5178,5179,606],{},[18,5180,605],{"href":585},[200,5182,5183,611],{},[18,5184,319],{"href":329},{"title":613,"searchDepth":614,"depth":615,"links":5186},[5187,5188,5195,5196,5197,5198,5199],{"id":28,"depth":614,"text":29},{"id":224,"depth":614,"text":225,"children":5189},[5190,5191,5192,5193,5194],{"id":229,"depth":615,"text":230},{"id":276,"depth":615,"text":277},{"id":283,"depth":615,"text":284},{"id":318,"depth":615,"text":319},{"id":333,"depth":615,"text":334},{"id":359,"depth":614,"text":360},{"id":447,"depth":614,"text":448},{"id":499,"depth":614,"text":500},{"id":531,"depth":614,"text":532},{"id":590,"depth":614,"text":591},[5201,5202,5203,5204,5205],{"q":635,"a":636},{"q":638,"a":639},{"q":641,"a":642},{"q":644,"a":645},{"q":647,"a":648},{},{"title":5,"description":631},{"id":5209,"title":5210,"author":6,"body":5211,"date":658,"description":5607,"extension":632,"faq":5608,"meta":5626,"navigation":650,"order":3225,"path":5627,"readTime":1789,"seo":5628,"stem":5629,"topic":656,"translationId":2450,"updated":658,"__hash__":5630},"guides\u002Fguides\u002Fclaude-code\u002Fbest-claude-model.md","Which Claude model is best? Opus vs Sonnet vs Haiku vs Fable",{"type":8,"value":5212,"toc":5594},[5213,5216,5221,5225,5290,5294,5382,5385,5408,5412,5415,5418,5438,5441,5445,5448,5455,5459,5462,5466,5469,5473,5476,5502,5513,5517,5520,5531,5534,5538,5543,5547,5550,5554,5557,5583],[11,5214,5215],{},"Anthropic now ships four current Claude models, and the right one depends less on which is \"smartest\" than on how hard the task is, how fast you need the answer and how many times you run it. This guide compares them on what matters in practice and ends with a simple way to choose.",[11,5217,5218,5219,827],{},"Model names, prices and limits below are from Anthropic's model documentation, checked on 12 October 2026. For the full price list, including caching and batch discounts, see ",[18,5220,4726],{"href":1675},[26,5222,5224],{"id":5223},"the-short-answer","The short answer",[34,5226,5227,5240],{},[37,5228,5229],{},[40,5230,5231,5234,5237],{},[43,5232,5233],{},"If you need…",[43,5235,5236],{},"Use",[43,5238,5239],{},"Why",[59,5241,5242,5254,5266,5278],{},[40,5243,5244,5247,5251],{},[64,5245,5246],{},"A strong default for almost everything",[64,5248,5249],{},[203,5250,83],{},[64,5252,5253],{},"Anthropic's default model: reasoning, coding and agents at a mid-range price",[40,5255,5256,5259,5263],{},[64,5257,5258],{},"The hardest problems, quality above cost",[64,5260,5261],{},[203,5262,66],{},[64,5264,5265],{},"The most capable widely released model",[40,5267,5268,5271,5275],{},[64,5269,5270],{},"Fast everyday coding, writing and analysis",[64,5272,5273],{},[203,5274,117],{},[64,5276,5277],{},"Half Opus's price, quicker answers",[40,5279,5280,5283,5287],{},[64,5281,5282],{},"High volume: classification, extraction, subagents",[64,5284,5285],{},[203,5286,163],{},[64,5288,5289],{},"A fraction of the cost, 1M context",[26,5291,5293],{"id":5292},"the-current-claude-models-compared","The current Claude models compared",[34,5295,5296,5314],{},[37,5297,5298],{},[40,5299,5300,5302,5305,5308,5311],{},[43,5301,45],{},[43,5303,5304],{},"API id",[43,5306,5307],{},"Context",[43,5309,5310],{},"Max output",[43,5312,5313],{},"Input \u002F output per 1M tokens",[59,5315,5316,5334,5350,5366],{},[40,5317,5318,5320,5325,5328,5331],{},[64,5319,66],{},[64,5321,5322],{},[300,5323,5324],{},"claude-fable-5-1",[64,5326,5327],{},"1M",[64,5329,5330],{},"128K",[64,5332,5333],{},"$10 \u002F $50",[40,5335,5336,5338,5343,5345,5347],{},[64,5337,83],{},[64,5339,5340],{},[300,5341,5342],{},"claude-opus-5-5",[64,5344,5327],{},[64,5346,5330],{},[64,5348,5349],{},"$4 \u002F $20",[40,5351,5352,5354,5359,5361,5363],{},[64,5353,117],{},[64,5355,5356],{},[300,5357,5358],{},"claude-sonnet-5-5",[64,5360,5327],{},[64,5362,5330],{},[64,5364,5365],{},"$2 \u002F $10",[40,5367,5368,5370,5375,5377,5379],{},[64,5369,163],{},[64,5371,5372],{},[300,5373,5374],{},"claude-haiku-5-5",[64,5376,5327],{},[64,5378,5330],{},[64,5380,5381],{},"$0.10 \u002F $0.50 *",[11,5383,5384],{},"* Haiku 5.5's price applies to prompts up to 100,000 tokens; above that it is $0.50 \u002F $2.50.",[11,5386,5387,5388,5391,5392,5395,5396,5399,5400,5403,5404,5407],{},"All four use tools and support adaptive thinking, where the model decides how long to think. Depth is set with an ",[203,5389,5390],{},"effort"," level from ",[300,5393,5394],{},"low"," to ",[300,5397,5398],{},"max",". Opus 5.5 and Haiku 5.5 default to ",[300,5401,5402],{},"medium","; Fable 5.1 and Sonnet 5.5 default to ",[300,5405,5406],{},"high",". Effort is often a bigger lever than model choice: a strong model at lower effort can beat a weaker model at high effort and cost less.",[26,5409,5411],{"id":5410},"claude-fable-51-the-most-capable","Claude Fable 5.1: the most capable",[11,5413,5414],{},"Fable 5.1 sits above Opus. It is built for the most demanding reasoning and long-horizon agent work: a large refactor across a codebase, a hard bug nobody has found, a research task that takes many steps.",[11,5416,5417],{},"Three things set it apart in practice:",[197,5419,5420,5426,5432],{},[200,5421,5422,5425],{},[203,5423,5424],{},"It always thinks."," Thinking cannot be switched off; you control depth with effort. A single hard request can run for many minutes.",[200,5427,5428,5431],{},[203,5429,5430],{},"It costs the most."," $10 in and $50 out per million tokens, two and a half times Opus 5.5.",[200,5433,5434,5437],{},[203,5435,5436],{},"It needs 30-day data retention."," It is not available under zero data retention unless Anthropic expressly authorises it. For companies with strict retention rules, that alone can rule it out.",[11,5439,5440],{},"Use it where a better answer is worth real money: architecture decisions, hard debugging, work you would otherwise give to your most senior engineer.",[26,5442,5444],{"id":5443},"claude-opus-55-the-default","Claude Opus 5.5: the default",[11,5446,5447],{},"Opus 5.5 is the model Anthropic recommends by default, and the right starting point for most teams. It is strong at multi-step reasoning, coding and agentic work, and at $4\u002F$20 it is cheaper per token than the Opus 5 it replaced ($5\u002F$25).",[11,5449,5450,5451,5454],{},"It also has a ",[203,5452,5453],{},"fast mode"," on Anthropic's own API: the same model at up to 2.5 times the output speed, for $8\u002F$40 per million tokens.",[26,5456,5458],{"id":5457},"claude-sonnet-55-the-everyday-model","Claude Sonnet 5.5: the everyday model",[11,5460,5461],{},"Sonnet 5.5 gives most of the capability at half Opus's price ($2\u002F$10) and answers faster. It is the sensible choice for everyday coding, writing, summaries and analysis where you notice speed and the task is not unusually hard, and for products that call the model many times per user.",[26,5463,5465],{"id":5464},"claude-haiku-55-high-volume-at-a-low-price","Claude Haiku 5.5: high volume at a low price",[11,5467,5468],{},"Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens: forty times cheaper than Opus 5.5 on input. It still has a 1M-token context window. Use it for classification, extraction, routing, tagging, summarising large volumes and as the worker model under a more capable orchestrator.",[26,5470,5472],{"id":5471},"which-claude-model-is-best-for-coding","Which Claude model is best for coding?",[11,5474,5475],{},"For coding, the answer depends on how the model is used:",[197,5477,5478,5484,5490,5496],{},[200,5479,5480,5483],{},[203,5481,5482],{},"Interactive coding in Claude Code or an IDE:"," Opus 5.5. It plans better and makes fewer wrong turns on multi-file changes, which saves more time than Sonnet's speed.",[200,5485,5486,5489],{},[203,5487,5488],{},"Long autonomous runs and the hardest bugs:"," Fable 5.1, if the budget allows. It is the model to escalate to when Opus gets stuck.",[200,5491,5492,5495],{},[203,5493,5494],{},"Quick edits, tests and boilerplate:"," Sonnet 5.5. Fast, cheap and good enough.",[200,5497,5498,5501],{},[203,5499,5500],{},"Subagents doing narrow jobs"," (searching a codebase, summarising files, running checks): Haiku 5.5 or Sonnet 5.5.",[11,5503,5504,5505,5509,5510,5512],{},"A common setup is Opus 5.5 for the main session and planning, with Sonnet or Haiku for subagents. Our ",[18,5506,5508],{"href":5507},"\u002Fen\u002Fguides\u002Fclaude-code\u002Fhow-to-use-claude-code","Claude Code guide"," shows how to switch models, and ",[18,5511,605],{"href":585}," covers what a day of agent work costs.",[26,5514,5516],{"id":5515},"opus-vs-sonnet-when-is-opus-worth-double","Opus vs Sonnet: when is Opus worth double?",[11,5518,5519],{},"Opus 5.5 costs twice what Sonnet 5.5 does per token. It earns that when:",[197,5521,5522,5525,5528],{},[200,5523,5524],{},"the task has many steps and an early mistake is expensive (agents, migrations, multi-file refactors);",[200,5526,5527],{},"the problem needs real reasoning, not recall (debugging, design trade-offs, analysis with conflicting data);",[200,5529,5530],{},"a wrong answer costs more than the tokens (legal, financial or customer-facing output that someone will act on).",[11,5532,5533],{},"For one-shot questions, drafting, summaries and high-volume pipelines, Sonnet 5.5 or Haiku 5.5 usually gives the same result for less. Measure on your own tasks: run twenty real examples through both and compare cost per correct result, not cost per request.",[26,5535,5537],{"id":5536},"which-models-you-get-on-each-claude-plan","Which models you get on each Claude plan",[11,5539,5540,5541,827],{},"In the Claude apps, the plan decides the models. Free gets Sonnet and Haiku. Pro and Team Standard get Opus, Sonnet and Haiku, with Fable only through paid usage credits. Max and Team Premium include Fable up to half of the weekly limit, and Enterprise bills every model at API rates. The full comparison is in ",[18,5542,1655],{"href":1654},[26,5544,5546],{"id":5545},"older-claude-models","Older Claude models",[11,5548,5549],{},"Anthropic still serves earlier models, including Claude Fable 5, Opus 5, Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 5, Sonnet 4.6 and Haiku 4.5. They are worth keeping only if you have a tested setup that depends on them. New work should start on the current generation, which is as cheap or cheaper: Opus 5.5 costs less than Opus 5, and Haiku 5.5 costs a tenth of Haiku 4.5 ($1\u002F$5) with five times the context.",[26,5551,5553],{"id":5552},"choosing-for-a-company-not-a-person","Choosing for a company, not a person",[11,5555,5556],{},"A team rarely needs one model. It needs the right model per task, and a way to stop everyone defaulting to the most expensive one:",[197,5558,5559,5565,5571,5577],{},[200,5560,5561,5564],{},[203,5562,5563],{},"Route by task."," Opus for planning and hard work, Sonnet for the everyday, Haiku for bulk jobs.",[200,5566,5567,5570],{},[203,5568,5569],{},"Set budgets per team or user",", so a long Fable session is a decision, not an accident.",[200,5572,5573,5576],{},[203,5574,5575],{},"Check data handling per model."," Fable 5.1's retention requirement, for example, matters if you have zero-retention commitments.",[200,5578,5579,5582],{},[203,5580,5581],{},"Keep the choice reversible."," Models change every few months; a gateway lets you switch without touching every integration.",[11,5584,5585,5586,5589,5590,827],{},"Walma AI Hub gives every employee Claude, GPT and other models through one gateway in your own EU region, with per-team model access, budgets and a full log. ",[18,5587,5588],{"href":580},"See how it works",", or read about ",[18,5591,5593],{"href":5592},"\u002Fen\u002Fai-hub\u002Fclaude","running Claude in the EU",{"title":613,"searchDepth":614,"depth":615,"links":5595},[5596,5597,5598,5599,5600,5601,5602,5603,5604,5605,5606],{"id":5223,"depth":614,"text":5224},{"id":5292,"depth":614,"text":5293},{"id":5410,"depth":614,"text":5411},{"id":5443,"depth":614,"text":5444},{"id":5457,"depth":614,"text":5458},{"id":5464,"depth":614,"text":5465},{"id":5471,"depth":614,"text":5472},{"id":5515,"depth":614,"text":5516},{"id":5536,"depth":614,"text":5537},{"id":5545,"depth":614,"text":5546},{"id":5552,"depth":614,"text":5553},"The best Claude model for coding, agents, writing and high-volume work in 2026: Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 compared on capability, speed, context and price per million tokens, with a simple way to choose.",[5609,5612,5614,5617,5620,5623],{"q":5610,"a":5611},"Which Claude model is best?","For most work, Claude Opus 5.5: it is Anthropic's default model, strong at reasoning, coding and agents, and costs $4 per million input tokens and $20 per million output tokens. Claude Fable 5.1 is the most capable model for the hardest problems, at $10\u002F$50. Claude Sonnet 5.5 is the fast, cheaper everyday model at $2\u002F$10, and Claude Haiku 5.5 is the high-volume model at $0.10\u002F$0.50.",{"q":5472,"a":5613},"Claude Opus 5.5 is the best default for coding and coding agents. Use Claude Fable 5.1 for the hardest, longest tasks (large refactors, difficult debugging, long autonomous runs) when quality matters more than cost. Use Claude Sonnet 5.5 for everyday coding where speed and price matter, and Claude Haiku 5.5 for subagents and simple, repetitive steps.",{"q":5615,"a":5616},"Is Opus better than Sonnet?","Opus 5.5 is the more capable model, particularly on long, multi-step reasoning and agentic work. Sonnet 5.5 is faster and half the price per token ($2\u002F$10 against $4\u002F$20), and is good enough for most everyday coding, writing and analysis. Many teams use Opus for planning and hard problems and Sonnet or Haiku for the bulk of the work.",{"q":5618,"a":5619},"What is Claude Fable?","Claude Fable 5.1 is Anthropic's most capable widely released model, above the Opus tier, for the most demanding reasoning and long-horizon agentic work. It always thinks before answering, can run for many minutes on one hard request, and costs $10 per million input tokens and $50 per million output tokens. It requires 30-day data retention, so it is not available under zero data retention unless Anthropic authorises it.",{"q":5621,"a":5622},"What is the cheapest Claude model?","Claude Haiku 5.5, at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens ($0.50\u002F$2.50 beyond that). It has a 1M-token context window and suits classification, extraction, routing and subagents.",{"q":5624,"a":5625},"How big is Claude's context window?","1 million tokens on every current model (Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5), with up to 128,000 output tokens per response. The older Claude Haiku 4.5 has 200,000 tokens.",{},"\u002Fguides\u002Fclaude-code\u002Fbest-claude-model",{"title":5210,"description":5607},"guides\u002Fclaude-code\u002Fbest-claude-model","quhy_TLKt3SkV7z3aEUiRyszjtxcvXym9NZh8L8RjBE",[5632],{"loc":5633,"href":1675},"en",1792405295984]