FastMCP

About

The FastMCP server provides a robust Python implementation of the Model Context Protocol (MCP), an open, JSON-RPC-based standard for AI clients to discover and invoke tools and resources. It offers both local stdio and remote Streamable HTTP transports, powered by a shared, centralized tool registry. Designed with production in mind, it prioritizes security through features like a sandboxed filesystem, SSRF guards, Origin validation, and bearer token authentication, ensuring safe and deterministic tool execution while keeping LLM reasoning within the agent.

Key Features

• Supports stdio and Streamable HTTP transports for flexible integration with AI agents.
• Centralized tool registry ensures consistent tool discovery and invocation across transports.
• Strong security model includes filesystem sandboxing and SSRF guards for HTTP client operations.
• Enforces strict input validation using Pydantic models and JSON Schema for all tools.
• Provides append-only artifact logging for observability, auditing, and multi-turn traceability.
• 0 GitHub stars

Use Cases

• Providing secure access to filesystem operations and HTTP fetching for AI workflows within a controlled environment.
• Enabling AI agents to safely discover and invoke deterministic tools like file I/O or HTTP requests via a standardized protocol.
• Building audit trails and managing ephemeral state for multi-turn agent interactions using artifact logging and an optional key-value store.