The Model Context Protocol (MCP)

What Is MCP

The Model Context Protocol is an open standard that connects AI models to external tools and data. Instead of hard-coding integrations, a model talks to MCP servers that expose databases, APIs, file systems and services.

How It Fits Together

The model runs inside an MCP client, usually the agent tool. The client connects to one or more MCP servers. Each server offers resources and tools, and the model can call them during a session.

Adding an MCP Server

claude mcp add database -- npx @example/database-mcp-server
claude mcp list

The numbered servers now expose tools like running SQL or reading schemas, which the model can invoke alongside normal file operations.

Example Tools

• Query a database and read live schemas.
• Search the codebase with IDE-scale indexing.
• Create and update issues in a tracker.
• Control a browser for testing and scraping.

Key Points

• MCP decouples models from hard-coded integrations.
• Servers expose tools and data to the agent.
• mcp add wires a server into Claude Code sessions.
• Use MCP to reach databases, browsers and APIs safely.