fast-agent-mcp

What it is and what it does

fast-agent-mcp is a framework for building and orchestrating LLM agents that interact with Model Context Protocol (MCP) servers. It provides both a CLI interface and a Python API, supporting agents as simple decorators or complex workflows. The framework handles agent skill management, MCP server connections, and multi-model provider support (Anthropic, OpenAI, Google, Azure, Ollama, and others via TensorZero), with built-in features for structured outputs, vision, PDF handling, and shell integration.

The package is designed for rapid agent development with minimal boilerplate. Agents can be run interactively via a prompt-toolkit-powered terminal interface with streaming responses via rich, or programmatically in automation and server modes. It includes diagnostic tools for MCP transport inspection, OAuth-based secret management via keyring, and support for agent chaining and workflow composition. The framework emphasizes declarative configuration through YAML files and simple Python decorators, enabling developers to focus on prompt and MCP server composition rather than infrastructure.

Use it for:

• Build a coding agent that uses MCP servers for file system access, LSP integration, and tool execution to assist with development tasks.
• Create a multi-step workflow where agents chain together (e.g., researcher → evaluator → optimizer) using shared MCP servers for data access.
• Set up an interactive CLI tool that lets users chat with an LLM agent backed by custom MCP servers for domain-specific tasks.
• Develop and test agent applications locally with model passthrough and playback LLMs before deploying to production.
• Expose an agent as an MCP server itself for use in other applications via stdio or HTTP transport with OAuth authentication.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

fast-agent-mcp is a CLI-first framework for building and running LLM agents with Model Context Protocol (MCP) server integration, supporting multiple LLM providers and interactive or programmatic execution.

Yes, if you are building LLM agents with MCP server integration. The framework is actively maintained, has low install friction, permissive licensing, and no known vulnerabilities. The large dependency footprint is justified by multi-provider support and feature completeness. Install only if you need agent orchestration; it is not suitable for simple LLM API calls.

Install

fast-agent-mcp on PyPI

pip install fast-agent-mcp

uv add fast-agent-mcp

poetry add fast-agent-mcp

Installing fast-agent-mcp

Before you install

Low install friction with a pure-Python wheel. Active maintenance with a release 1 day old. The package has 37 runtime dependencies including fastapi, anthropic, openai, and google-genai, which is a substantial dependency footprint but typical for a multi-provider LLM framework.

License in practice

Apache License 2.0 permits commercial and derivative use with attribution and patent protection. The license is permissive and widely compatible with most projects.

Quickstart

pip install fast-agent-mcp

from fast_agent import FastAgent
import asyncio

fast = FastAgent("Agent Example")

@fast.agent(instruction="Given an object, respond only with an estimate of its size.")
async def main():
    async with fast.run() as agent:
        await agent.interactive()

if __name__ == "__main__":
    asyncio.run(main())

Requires Python 3.12 or later (supports up to <3.15). An LLM provider API key (Anthropic, OpenAI, Google, or compatible) must be configured via environment or keyring to run agents.

Verify before relying

• Whether all 37 runtime dependencies are required for basic usage or if subsets can be installed for specific providers
• Performance characteristics and latency overhead of MCP transport diagnostics and Streamable HTTP support
• Compatibility matrix with specific MCP server versions and transport types (stdio vs. HTTP)

Package facts

Evidence: fast_agent_mcp-0.10.7-py3-none-any.whl

Tags

More Artificial Intelligence packages

Further reading