The best MCP servers on GitHub

An MCP server is a small program that exposes tools, data or prompts to an AI assistant through the Model Context Protocol, an open standard. The assistant (the MCP client) calls the server's tools, for example to query a database, open a GitHub issue or browse a web page.

Local servers run as a command such as npx -y <package>, uvx <package> or docker run. Add that command to claude_desktop_config.json for Claude Desktop, run claude mcp add for Claude Code, or put it in .cursor/mcp.json (Cursor) or .vscode/mcp.json (VS Code). Remote servers only need their URL. Every server page on appsgit has ready-made snippets for all four clients.

Every server listed here is open source, so running it costs nothing. Many servers wrap a commercial API, though, and need an API key for that service, which may be paid. Pages flag those servers with a Needs API key badge and list the environment variables to set.

Official means the server is published by the vendor itself: either under a verified domain namespace in the official MCP registry (such as com.stripe/...), or from a GitHub organization with a verified domain. 231 servers on appsgit are official.

Servers come from the official MCP registry plus GitHub repositories tagged mcp-server with 200+ stars. A server is listed only if it has a GitHub repository, a recognised open-source license, a real description, a commit in the last six months, an install path (npm, PyPI, Docker or NuGet package, or a remote URL) and either 100+ stars or an official publisher.

By GitHub stars, the most popular MCP servers listed on appsgit are Browser Use, Gemini CLI, Worldmonitor. See the trending list for the servers gaining stars fastest right now.

A local MCP server runs with your user's permissions and an assistant can call any tool it exposes, so treat it like any other software you install: prefer official or well-maintained servers, read the README, give API tokens the narrowest scope possible and review tool calls before approving them.