Connecting MCP Servers to Claude Code

So far Claude Code has worked with your files and your terminal. MCP — the Model Context Protocol — lets it reach further: into your issue tracker, your database, your monitoring dashboard, and hundreds of other tools.

MCP is an open standard for connecting AI tools to external systems. You connect a server (for a specific tool, like GitHub or Sentry), and Claude can then read and act on that system directly — instead of you copy-pasting data into the chat.

You add servers from your terminal (not inside a session) with the claude mcp command. Most cloud tools expose a remote HTTP server:

claude mcp add --transport http sentry https://mcp.sentry.dev/mcp

Some tools run as a local process on your machine instead. Those use the stdio transport, with the command after a -- separator:

claude mcp add --transport stdio playwright -- npx -y @playwright/mcp@latest

A scope decides which projects a server loads in, and whether your team gets it too:

• local (the default) — only you, only in this project. Stored in ~/.claude.json.
• project — shared with your team via a .mcp.json file at the project root that you commit to version control.
• user — only you, but across all your projects.

claude mcp add --transport http paypal --scope project https://mcp.paypal.com/mcp

From the terminal you can claude mcp list, claude mcp get <name>, and claude mcp remove <name>. Inside a session, the /mcp command shows each server's status and tool count, and is where you complete login for servers that need OAuth:

/mcp

Once a server is connected, just ask in plain language — "What are the most common Sentry errors in the last 24 hours?" — and Claude uses the server's tools. Some servers also expose data you can pull in with an @ mention, like @github:issue://123.