MCP — connecting external tools
Until now Claude Code has only worked with what's already built in: reading and editing files, running shell commands, searching code. But the team's real information — an open ticket in an issue tracker, a production error in a monitoring tool, a document in an external system — lives outside the r
MCP is like a universal travel power adapter: instead of a custom cable for every country's socket, one fixed adapter connects your device to whatever socket you find.
- MCP (Model Context Protocol)
- An open standard that defines how an AI agent connects to external programs and gains new tools and data from them, without changing a single line in the agent's own code.
- MCP server
- An external program that exposes tools and data according to the MCP standard — for example a server that talks to GitHub, Sentry, or a database.
- MCP client
- The application that connects to MCP servers and uses the tools they expose — here, Claude Code itself.
- connected tool
- A new tool that becomes available to the agent the moment the MCP server providing it is configured and connected — for example mcp__github__issue_read.
