//
Search across all documentation pages
8 pages in this section.
Learn how Model Context Protocol (MCP) standardizes AI tool discovery and invocation with JSON-RPC 2.0, replacing one-off connectors.
Connect an existing MCP server to an agent client. Learn to list tools and call a tool from a thin Python client, then wire it into an agent loop.
Learn about MCP's core primitives: tools, resources, and prompts. Understand their uses, how models and hosts interact with them, and when to apply each.
Compare MCP's transport options: stdio for local, Streamable HTTP for remote, and legacy HTTP+SSE for older patterns. Learn how transport choice impacts security, latency
Explore the MCP server ecosystem with over 10,000 servers in production. This cheatsheet maps categories, evaluation axes, and adoption patterns.
Learn ten best practices for evaluating, scoping, connecting, and operating MCP servers to provide agents with tools safely and securely.