The best MCP servers in 2026: top picks by use case

There are thousands of MCP servers. Most people need about ten. This list starts with the most popular servers on GitHub, then groups the best ones by what they connect to, with install commands and the notes that matter when you approve a server for more than one person.

Repositories and star counts were checked on GitHub on 12 October 2026. For a searchable list of 140+ business tools and whether each has an official server, see our MCP directory.

The top 10 MCP servers at a glance

Ranked by GitHub stars, which measure attention rather than quality.

Anthropic's own reference servers (about 91,000 stars as a collection) are worth knowing too: everything, fetch, filesystem, git, memory, sequential thinking and time. They are examples of the protocol more than production connectors, so use them for learning and for local work.

How we picked

Two rules of thumb shaped the list beyond popularity. Prefer servers hosted by the vendor of the underlying system, because they use real OAuth and cannot be typosquatted. And read the tool list before approving: the best server is the one whose tools match what your team should be able to do, not the one with the most tools. If you have not seen what an MCP server is, start there.

Best MCP servers for coding

These are the servers most worth adding to Claude Code, Cursor or Codex. In Claude Code:

claude mcp add playwright npx @playwright/mcp@latest
claude mcp add-json github '{"type":"http","url":"https://api.githubcopilot.com/mcp","headers":{"Authorization":"Bearer YOUR_GITHUB_PAT"}}'

Context7 sets itself up with npx ctx7 setup, or connects remotely at https://mcp.context7.com/mcp with an API key. General setup per client is in our guides for Claude Code, Cursor and Claude Desktop.

GitHub (guide). The official server covers repositories, issues, pull requests, code search, actions and security alerts. Remote, OAuth, hosted by GitHub. It is the single most used server we see. Watch the scopes: the default grants more than an agent that only reads issues needs, and the 2025 prompt-injection demonstration used exactly this server. Give it a token scoped to the repositories the agent works on.

GitLab. Equivalent coverage for GitLab projects, merge requests and pipelines. Self-hosted GitLab instances can run the server inside the same network, which is the right answer for regulated environments.

Azure DevOps. Work items, repos, pipelines and boards. Widely used in Microsoft-centric Nordic and German companies. Runs against your organisation with a PAT or Entra identity.

Sentry (guide). Issues, stack traces, releases. Very useful for "explain this error and propose a fix" workflows in Claude Code. Read-only by nature, low risk.

Context7. Not a system connector but a documentation source: it feeds up-to-date library docs to the model so it stops hallucinating APIs. The most-starred MCP server of all. Low risk, high value for code quality.

Serena. Gives the agent semantic code tools (find symbol, find references, edit by symbol) instead of plain text search. Worth it in large codebases, where it saves the agent reading whole files.

Datadog and Grafana (Datadog guide, Grafana guide). Metrics, logs and dashboards for "why did this alert fire" questions. Read-only by default, which is how it should stay.

Best MCP servers for project management and docs

Atlassian (Jira and Confluence) (guide). The official remote server covers both. Jira tools include creating and transitioning issues; hide the destructive ones for most users. Confluence access is the classic source of untrusted content, so pair it with read-only sessions.

Linear (guide). Clean, fast, remote, OAuth. Tools map closely to how teams actually use Linear. A good first server for product teams.

Notion (guide). Pages, databases and search. Notion workspaces mix internal and external content, so treat results as untrusted.

Slack. Reading channels and posting messages. Posting is a write action with real consequences; require approval or restrict to specific channels.

Best MCP servers for data

Snowflake, Postgres, BigQuery, Databricks (Snowflake guide). Each has a server, official or well-maintained, and Google's open-source MCP Toolbox for Databases covers many of them from one server. The rule for all of them: a read-only database role, an allowlist of schemas, and a row limit. A database server with a read-write connection string is the highest-risk thing on this page.

Filesystem. The reference local server. Start it with an explicit list of allowed directories and nothing else. Never on the home directory.

Google Drive and SharePoint. Document search and reading. Both are large pools of untrusted content and often contain personal data; think about where the server runs and what leaves the region.

Best MCP servers for browsers and testing

Playwright (guide). Microsoft's server lets the agent drive a real browser: navigate, click, fill forms, take screenshots, run tests. Excellent for QA and for agents that need to check a web app. The browser sees whatever the agent visits, so it is also an injection vector; run it in an isolated profile without logged-in sessions.

Chrome DevTools. Google's server for performance traces, network inspection and debugging in a live Chrome. Useful for front-end teams, and second only to Context7 in stars.

Best MCP servers for web search and scraping

Firecrawl, Exa and Brave Search. Firecrawl turns pages into clean text and crawls sites; Exa and Brave give the model a search engine. Everything they return is content someone else wrote, so treat it as untrusted input, especially in sessions that also hold write tools.

Best MCP servers for design

Figma. Figma's own Dev Mode server and the community Figma Context MCP both give coding agents layout, spacing and component data, so the code matches the design rather than a screenshot of it.

Blender MCP. A community server that lets Claude build and change 3D scenes in Blender. Popular, fun and a good demonstration of what MCP can drive; not a business connector.

Best MCP servers for infrastructure and cloud

Azure. Microsoft's servers (catalogued in microsoft/mcp) cover a wide range of Azure services. Scope it to the subscriptions and resource groups the agent needs. For EU companies running on Azure it is the natural way to give an agent operational visibility without handing out portal access.

AWS. Amazon publishes a family of servers per service rather than one. Same scoping advice.

Cloudflare, Supabase and Terraform. Each vendor runs an official server. Supabase in particular can run SQL against your project; connect it to a development project, not production.

Docker and Kubernetes. Community and vendor servers exist for both. Anything that can run containers or apply manifests is effectively remote code execution; require approval per call.

Best MCP servers for business systems

n8n and Zapier. Both expose their workflows as MCP tools, which turns any automation your ops team has already built into something an agent can trigger. Powerful and a wide blast radius; approve per workflow.

Stripe (guide). Payments, customers, invoices. Read tools are fine for support agents; write tools should be behind approval and probably a separate, restricted key.

CRM and sales. HubSpot (guide), Salesforce (guide) and Gong (guide) all have official servers. Start read-only.

Marketing and analytics. Google Analytics, Search Console, Google Ads, Tag Manager and Ahrefs all have servers; our best MCP servers for marketing and SEO covers them in one place.

ERP and finance systems. Fortnox, Visma, SAP, DATEV and Dynamics have official or community servers of varying maturity. This is the category where most companies end up writing their own narrow server instead: read-only, a handful of tools, hosted inside the tenant. Our MCP gateway guide explains why.

How to evaluate a server before approving it

1. Who runs it? Vendor, well-known maintainer, or an unknown package? Prefer the first.
2. Remote or local? Remote with OAuth is easier to govern. Local needs version pinning and directory restrictions.
3. What are the tools? Read the list. Count the write, send and delete tools. Decide who gets them.
4. What do the descriptions say? Descriptions that reference other tools, files outside the server's purpose, or secrecy are disqualifying.
5. Where does it run? For remote servers, which region. For EU data, this decides the answer.
6. What will it return? If results can contain content others control, plan for injection: read-only sessions, result inspection, approval on writes.

Running them together

Ten servers across fifty developers and three clients is where the list stops being the hard part. The hard part becomes who is allowed which tools, where the credentials live, and what the log says. That is the job of a gateway.

Walma AI Hub hosts approved servers, including internal ones over ERP and CRM systems, inside the customer's Azure tenant in an EU region, and applies one policy and one log across Claude, GPT, Codex and Cursor. If you are choosing your first ten servers, we are happy to walk through the list with you.