AgentChat is an LLM-based agent communication platform with built-in default Agent and supports user-defined Agent. Through multiple rounds of dialogue and task collaboration, the Agent can understand and assist in completing complex tasks. The project integrates technologies such as LangChain, Function Call, MCP protocol, RAG, Memory, Milvus and ElasticSearch to achieve efficient knowledge retrieval and tool invocation, and uses FastAPI to build high-performance back-end services.