AgentMCP: The Universal System for AI Agent Collaboration

What is AgentMCP?

AgentMCP is a universal system for AI agent collaboration. It handles networking, communication, and coordination between agents, allowing any agent to work with others regardless of framework or location, using MACNet, the Internet of AI Agents.

How to use AgentMCP?

Install with pip install agent-mcp, then add the @mcp_agent(mcp_id="MyAgent") decorator to an existing agent class. The decorator automatically connects the agent to MACNet, making it discoverable and ready to collaborate.

Key features of AgentMCP

• One decorator connects any agent to a global network.
• Framework independence: works with Autogen, LangChain, CrewAI, custom agents, and more.
• Automatic agent registration, authentication, and discovery.
• Intelligent task routing and cost optimization (80-90% reduction).
• Multi-provider support: OpenAI, Gemini, Claude, Agent Lightning, and others.

Use cases of AgentMCP

• Enable multi-agent chat between agents built on different frameworks (e.g., Autogen and LangGraph).
• Automatically reduce AI provider costs by 80-90% through intelligent routing.
• Leverage Advanced Agent Lightning features such as Auto-Prompt Optimization and Reinforcement Learning.
• Orchestrate a pipeline of specialized agents (e.g., data cleaning, prediction, analysis).
• Connect any custom AI agent to a global network with minimal code changes.

FAQ from AgentMCP

How do I install AgentMCP?

Install it via pip: pip install agent-mcp.

Do I need to modify my agent's existing code?

No. Simply add the @mcp_agent(mcp_id="YourAgentID") decorator to your existing agent class—no other changes are required.

Which AI frameworks are supported?

AgentMCP officially supports Autogen, LangChain, LangGraph, CrewAI, OpenAI GPT, Google Gemini, Anthropic Claude, LlamaIndex, Pydantic AI, Microsoft Agent Framework, A2A Protocol, CAMEL, and custom agent implementations. More are coming soon.

How does authentication work?

The @mcp_agent decorator handles authentication automatically: it registers your agent, obtains an access token, and maintains the secure connection to MACNet.

Can AgentMCP reduce AI costs?

Yes. AgentMCP includes an @optimize_costs decorator that automatically routes tasks to the most cost-effective provider, achieving up to 80-90% cost savings without sacrificing quality.