MCP server configuration in Claude Code

MCP (Model Context Protocol) is an open-source standard that lets Claude Code connect to external systems — databases, APIs, file systems, browsers, and other tools — through a consistent interface. When you configure an MCP server in Claude Code, you're telling Claude how to launch or connect to a process that exposes external capabilities as callable tools. Claude can then invoke those tools during a session to read files, query databases, open pull requests, or retrieve live metrics without the developer switching context. MCP servers can run in three ways: as local processes communicating over standard input/output (stdio), as remote services reachable over HTTP (using the streamable-http transport, formerly called SSE), or embedded directly in SDK applications. Each server definition lives in a configuration file — either a project-level .mcp.json committed to version control, or a personal ~/.claude.json entry scoped to your machine. Claude Code reads these files on startup and makes each server's tools available in the session. Configuration is managed through either the CLI command `claude mcp add` or by hand-editing the relevant JSON file. Active connections can be inspected at any time with the `/mcp` slash command inside a Claude Code session. Administrators on Team and Enterprise plans can also push and enforce MCP configurations org-wide through managed policy settings.