AgentChat is an LLM-based intelligent agent communication platform that supports custom agent creation alongside built-in default agents. Through multi-turn dialogue and task collaboration, agents can understand and assist with complex tasks. The project integrates LangChain, Function Calling, MCP protocol, RAG, Memory systems, Human-in-the-loop (HITL), Skills, Milvus vector database, and Elasticsearch to enable efficient knowledge retrieval and tool invocation, using FastAPI for high-performance backend services.