Skip to content

MCP Server Directory

39 servers|6 categories
MCP Specification
39 servers
Sort:

Category Breakdown

Official MCP Servers Repo
Official12
Developer Tools6
AI & LLM4
Data & APIs6
Productivity5
Community6
Data as of July 19, 2026

What This Tool Does

Production-ready MCP servers are easiest to identify by maintainer and maintenance status. This directory tracks 39 Model Context Protocol servers across 6 categories with install commands and GitHub activity. Official reference servers from the Model Context Protocol project and vendor-owned servers such as Cloudflare, Stripe, and Supabase are the safest starting points; 8 early reference servers are now archived upstream and should not be used for new deployments.

Use How to Use for execution steps and FAQ for constraints, policies, and edge cases.

Last updated:

This tool is provided as-is for convenience. Output should be verified before use in any production or critical context.

Agent Invocation

Best Path For Builders

Browser workflow

Runs instantly in the browser with private local processing and copy/export-ready output.

Browser Workflow

This tool is optimized for instant in-browser execution with local data handling. Run it here and copy/export the output directly.

/mcp-directory/

For automation planning, fetch the canonical contract at /api/tool/mcp-directory.json.

How to Use MCP Server Directory

  1. 1

    Browse by category

    Filter MCP servers by category (databases, APIs, productivity, file systems). Each category shows popular servers with install commands and descriptions.

  2. 2

    Copy install command

    Use the Copy control next to any server to copy the exact installation command (npm, pip, or cargo). Run it in your terminal.

  3. 3

    Check requirements and setup

    Read the setup instructions for each server—many require environment variables, API keys, or specific runtime versions. Check before installing.

  4. 4

    Find examples and documentation

    Each server listing includes links to GitHub repos and official docs. Review example usage before integrating into your MCP client.

  5. 5

    Search and filter by feature

    Use the search bar to find servers by name or capability (e.g., 'search' to find web search servers, 'database' for DB connectors).

Frequently Asked Questions

What is MCP Server Directory?
MCP Server Directory is a curated, searchable catalog of Model Context Protocol servers. It helps developers discover MCP servers for Claude and other AI assistants, complete with install commands and categories.
How do I use MCP Server Directory?
Browse or search the directory by name, category, or functionality. Each server entry includes a description, installation command you can copy directly, and links to the source repository.
Is MCP Server Directory free?
Yes. This tool is free to use with immediate access—no account required.
Does MCP Server Directory store or send my data?
No. All processing happens entirely in your browser. Your data never leaves your device — nothing is sent to any server.
What is the Model Context Protocol (MCP)?
MCP is an open protocol that lets AI assistants like Claude connect to external tools and data sources. MCP servers provide capabilities such as file access, database queries, API integrations, and more — extending what AI agents can do.

Which MCP servers are production ready?

There is no official certification for MCP servers, so "production ready" comes down to verifiable signals: who maintains the server, whether it is actively maintained or archived, how widely it is deployed, and whether its permission scope fits your threat model. The table below lists the 31 actively maintained servers in this directory, sorted by GitHub stars. Star counts refresh weekly; data as of July 19, 2026.

Actively maintained MCP servers (31)

Server Maintainer Category Transport GitHub Stars
Filesystem Model Context Protocol Official stdio 88.6k
Memory Model Context Protocol Official stdio 88.6k
Fetch Model Context Protocol Official stdio 88.6k
Sequential Thinking Model Context Protocol Official stdio 88.6k
Git Model Context Protocol Developer Tools stdio 88.6k
Everything Model Context Protocol Community stdio 88.6k
Context7 Upstash AI & LLM stdio 59.4k
Chrome DevTools ChromeDevTools Developer Tools stdio 47.2k
Playwright Microsoft Data & APIs stdio 35.2k
GitHub GitHub Official stdio 31.5k
FastMCP Prefect AI & LLM stdio 26.3k
Serena Oraios AI Community stdio 22.9k
n8n Workflows Romuald Członkowski Data & APIs stdio 22.3k
Figma Context Graham Lipsman AI & LLM stdio 15.4k
Firecrawl Firecrawl Data & APIs stdio 7.0k
Notion Notion Productivity stdio 4.5k
Exa Search Exa Labs Community stdio 4.2k
Obsidian MarkusPfundstein Productivity stdio 4.1k
Cloudflare Cloudflare Data & APIs stdio 4.0k
Supabase Supabase Data & APIs stdio 2.8k
Stripe Stripe Data & APIs stdio 1.7k
MySQL Ben Borla Community stdio 1.5k
Kubernetes Suyog Sonwalkar Developer Tools stdio 1.5k
Google Calendar nspady Productivity stdio 1.1k
Sentry Sentry Developer Tools stdio 781
Docker Christian Kreiling Developer Tools stdio 731
Todoist Abhiram Nair Productivity stdio 391
Linear Jeremy Hadfield Developer Tools stdio 346
AWS MCP Servers AWS Samples Community stdio 232
LangChain Tools Andrew Wason AI & LLM stdio 205
Raygun Raygun Community stdio 20

Reference servers maintained by the Model Context Protocol project share the star count of the modelcontextprotocol/servers monorepo. Source: each server's GitHub repository, refreshed July 19, 2026.

How do I evaluate an MCP server before production use?

  1. Check the maintainer. Servers owned by the vendor of the underlying service (GitHub, Cloudflare, Stripe, Supabase, Microsoft's Playwright) track API changes fastest.
  2. Check maintenance status. Several early reference servers are archived and no longer receive fixes — archived servers are listed below and should be treated as end-of-life.
  3. Review the permission scope. A filesystem or database server should be configured with the narrowest possible access (allowed directories, read-only credentials) before an agent can call it.
  4. Prefer stdio transport for local, single-user setups and audit any server that makes outbound network calls with your data.
  5. Pin versions. Install commands like npx -y fetch the latest release; pin an exact version in production so server behavior does not change under your agent.

Which MCP servers are archived?

These 8 servers are archived upstream. They may still work today, but they no longer receive security fixes or API updates, so do not build new production deployments on them:

PostgreSQL, SQLite, Brave Search, Google Maps, Slack, Puppeteer, EverArt, Google Drive.