MCP Server Directory

MCP (Model Context Protocol) is an open standard that lets AI agents connect to external tools and data sources. An MCP server exposes a specific capability — like querying a database, browsing the web, or reading files — that your AI agent can call during a task. Think of it as a plugin system for AI agents.

Every server in this directory goes through an 8-point review covering authentication handling, dependency risk, known CVEs, data exfiltration potential, and maintenance activity. Servers earn a "Verified", "Caution", or "Untested" rating. The full methodology is documented at codaone.ai/mcp/security.

Most servers here are compatible with Claude Code, Cursor, Windsurf, VS Code (via the MCP extension), and any client that implements the MCP spec. Each server's detail page lists which clients have been tested and confirmed working.

Open any server's detail page and you'll find copy-paste install commands for each supported client. For Claude Code, that's typically a one-line claude mcp add command. The Config Generator can also build a ready-to-use config file if you want multiple servers at once.

The directory itself is free. The servers listed are mostly open-source and free to self-host. Some connect to paid third-party services (like Stripe or Salesforce APIs), so you may need your own API keys for those. Pricing details are noted on each server's page.

A regular API requires you to write code that calls it explicitly. An MCP server exposes the same functionality in a format AI agents can discover and invoke autonomously — the agent decides when and how to use it based on the task at hand. No manual API call code required on your end.

Yes. Each server goes through the same security review before listing. To submit, open an issue or pull request on the directory's GitHub repo (linked in the footer). Include a README, license, and any existing audit or security disclosure information you have.