mcp-agent

Here's a summary of the README content, maintaining the original structure and technical details:

mcp-agent is a composable framework for building AI agents using the Model Context Protocol (MCP). Inspired by Anthropic's work on effective agents, it simplifies MCP server connection management and implements composable patterns for building production-ready AI agents. It also includes OpenAI's Swarm pattern for model-agnostic multi-agent orchestration.

Key features include: MCP server management, Agents with access to MCP servers, and AugmentedLLMs enhanced with tools from MCP servers. It provides implementations for patterns like Parallel, Router, IntentClassifier, Evaluator-Optimizer, Orchestrator-Workers, and Swarm.

mcp-agent supports various use cases, including integration with Claude Desktop, Streamlit applications (Gmail agent, RAG chatbot), Marimo notebooks, and Python scripts. It offers advanced features like composability, signaling, human input, and app configuration. The project welcomes contributions and has a roadmap focused on durable execution, memory, streaming, and expanded MCP capabilities.