//
Search across all documentation pages
10 pages in this section.
Learn how the Model Context Protocol (MCP) connects AI models to external tools, data, and services using three core primitives.
Learn to define a single MCP tool and connect a client over stdio using the Python SDK. Examples cover installation, tool definition, and server setup.
Define callable tools in an MCP server. Learn to write Python tool schemas and handlers for models to discover, invoke, and interpret reliably.
Define MCP resources using the @mcp.resource() decorator with URIs or URI templates. Learn to expose files, database rows, or API responses as addressable MCP resource URIs.
Compare MCP tools, resources, and prompts. Learn to choose the right primitive for server capabilities like fetching files, ensuring predictable client interactions.
Design predictable MCP servers with this checklist. Learn best practices for tool schemas, resource URIs, prompt templates, and server deployment.
A single-page roundup of every highlight bullet from the 9 pages in the MCP Concepts section, grouped by source page so you can scan all 39 takeaways without opening each article individually.