mcp-agent

mcp-agent is an open-source Python framework from LastMile AI for building composable, model-agnostic agents on the Model Context Protocol using proven workflow patterns.

mcp-agent, from LastMile AI, is a lightweight framework that makes it practical to build agents on top of the Model Context Protocol. It introduces an AugmentedLLM, an LLM enhanced with tools from a collection of MCP servers, and implements every workflow pattern from Anthropic's Building Effective Agents plus the OpenAI Swarm pattern. Because each pattern is itself an AugmentedLLM, patterns compose and chain cleanly, keeping agent logic modular and model-agnostic. The framework provides full MCP client support, including tools, resources, prompts and notifications, plus advanced features like OAuth, sampling, elicitation and roots, and it supports persistent state through durable execution. MCP is low-level, and mcp-agent handles the mechanics of server connections, LLM interaction and human-in-the-loop signals so developers can focus on business logic. Documentation, an SDK and CLI are available at docs.mcp-agent.com, making it a strong open-source choice for teams standardizing on MCP.