Features · Extend it · since version 0.2.21
MCP servers
Connect Claude Code to outside tools, databases and services through the Model Context Protocol.
What it is
MCP (Model Context Protocol) is an open standard for giving an AI assistant extra tools. An MCP server might expose your issue tracker, a database or a design tool. Once you add a server, Claude can call its tools directly instead of you copying data into the chat. Servers run either on your machine (started as a local command) or remotely over HTTP.
Why it matters
- Work with your real toolsClaude can read a ticket, query a database or check monitoring data itself, without copy and paste.
- Share a setup with the teamServers added at project scope are saved in a .mcp.json file you can commit to the repository.
- Control what is connectedThe /mcp command shows each server's status and lets you reconnect, sign in, enable or disable it.
How it works
- Add a remote server with claude mcp add --transport http <name> <url>, or a local one with claude mcp add <name> -- <command>.
- Pick a scope: local (the default, just you in this project), project (saved to .mcp.json) or user (all your projects).
- Servers from a project's .mcp.json wait for your approval before they run.
- In a session, use /mcp to check status and to sign in to servers that need OAuth.
- Only connect servers you trust: what a server returns can try to steer Claude.
Hands-on example coming
This feature has no reviewed example yet. The official documentation below has the details.
Update history
Version 0.2.32 · Mar 5, 2025
Updated claude mcp add gained a step-by-step setup.
Version 0.2.21 · by Mar 4, 2025
Introduced Already part of Claude Code when the changelog began with this version.