RepoDepot
Category

MCP Servers

Model Context Protocol servers let agents reach beyond the chat — search your docs, query databases, run code, control your tools. MCP is an open standard, so the same server works across Claude Code, Cursor, Codex, and other MCP clients. Every server listed is verified, with live GitHub data and a plain-English pitch of what it actually does.

Frequently asked

About mcp servers

What is an MCP server?+

A Model Context Protocol (MCP) server exposes tools, data, and prompts to an AI agent through a standard protocol. The server can run locally or remotely; the agent calls its tools as if they were native capabilities.

Which agents can use MCP servers?+

MCP is an open standard. Claude Code, Cursor, Codex, and a growing list of other agents can all connect to the same MCP server. One implementation, many clients.

How do I install an MCP server?+

Most MCP servers ship as a Node, Python, or Docker package. Add the server to your agent’s MCP config (usually a JSON file with the command, args, and env). Some Claude Code plugins install the server automatically.

Are MCP servers secure?+

Treat any MCP server like any other dependency that runs with your user permissions. Audit the source for popular ones; for unknown ones, run them in a container or sandbox. The MCP spec defines transport (stdio or SSE) but not sandboxing.

Browse other categories