Turn standard Python functions into tools for LLMs
Forget complex protocol configuration and attach decorators to your usual Python functions to create tools that LLMs can call. FastMCP is a framework for building Model Context Protocol (MCP) servers and clients. Developers focus on business logic while the framework handles details like communication and validation.
Function declaration as tool definition
FastMCP converts Python functions into MCP-compliant tools. Attaching a decorator to a function automatically generates the schema, input validation, and documentation. When creating a server, simply connecting a URL automatically handles transport negotiation, authentication, and protocol lifecycle management. This allows developers to work without knowing the details of the MCP protocol.
Three core axes: server, client, and app
FastMCP provides three core elements: server, client, and app. The server wraps Python functions into tools, resources, and prompts that LLMs can understand. The client connects to any local or remote MCP server with full protocol support. The app feature renders interactive UIs directly within the chat window, making tools more intuitive to use. An official counterpart version maintained by the same team exists for TypeScript developers.
Installation and documentation access
To add FastMCP to a project, run the uv add fastmcp command. Guides are provided for upgrading from previous versions (2, 3, or the low-level SDK). The official documentation site is available at gofastmcp.com, and it also supports documentation in the llms.txt format, which is easy for LLMs to consume. This format can serve as a sitemap for the documentation or include the entire documentation to fit within an LLM's context window.
Scalability and governance
FastMCP handles the application layer and is used alongside Prefect Horizon in enterprise environments. Horizon is an enterprise MCP gateway that scales the distribution of servers and tools across teams and manages deployment, discovery, security, and usage through centralized governance. When deploying FastMCP servers on GitHub, you can use branch previews and instant rollback features. Access is protected with SSO and tool-level RBAC, and audit logs and observability are provided to control the entire MCP stack.
Pre-implementation checks
FastMCP is a framework that standardizes MCP-based LLM tool development and is the most widely used choice in the Python ecosystem. For TypeScript-based projects, you should check the official counterpart library. If large-scale deployment and security governance are needed in an enterprise environment, consider integration with Horizon. The uv package manager is recommended for installation, and existing MCP SDK users must check the version upgrade guide.
