AgentChat

What does this agent do, and when should you use it?

AgentChat is a large-language-model conversation system with a FastAPI backend and a Vue 3 plus Element Plus frontend. It includes default agents and user-defined agents, with multi-turn conversations, sub-agent collaboration, task workflows, and chained tool calls. Its knowledge-base features support multiple file formats, semantic chunking, vector retrieval, and RAG question answering. The platform can dynamically load MCP servers and use a human-in-the-loop flow to turn OpenAPI information into MCP servers. It can run through Docker Compose or as separate Python and Node.js services, with MySQL, Redis, and configurable Milvus, ChromaDB, and Elasticsearch components in its deployment and retrieval stack.

A user interacts with an agent through the Vue frontend. The system keeps short-term context within 3,000 tokens, summarizes older history, and persists long-term information such as user preferences. Agents can make ordered, dependency-driven tool calls—for example, passing results from A to B to C—and the project provides more than 10 built-in tools plus custom tools created by uploading Swagger/OpenAPI definitions. For knowledge-base questions, it semantically chunks supported content formats, retrieves vectors, and uses the retrieved material for RAG answers. Its HITL workflow converses over OpenAPI information, requests human confirmation at key decisions, and produces MCP server-generation results. FastAPI serves the backend, and Swagger documentation is available at /docs after the backend starts.

1. A team building an internal Q&A service can add source material to the knowledge base and use semantic retrieval with RAG for answers.
2. A business team that needs to connect several APIs into one task can upload Swagger/OpenAPI definitions and let an agent invoke tools in dependency order.
3. A developer who must review generated integrations can intervene at key decisions while an OpenAPI definition is converted into an MCP server.
4. An operations user managing multiple conversational applications can switch between the workspace and application center and filter call counts and token usage by agent, model, and time range.
5. An engineering team deploying a private conversational service can use Docker Compose or run the FastAPI backend and Vue frontend separately.

What are this agent's strengths and limitations?

How do you install or deploy this agent?

The documented runtime requirements are Python 3.12+, Node.js 18+, MySQL 8.0+, Redis 7.0+, and Docker 20.10+. For Docker: git clone https://github.com/Shy2593666979/AgentChat.git, cd AgentChat, edit docker/docker_config.yaml, then run cd docker and docker-compose up --build -d. For local setup, in src/backend, run pip install -r requirements.txt, or install uv and run uv sync; in src/frontend, run npm install and npm run dev. The supplied documentation does not specify model-provider credential fields, MySQL/Redis initialization, or a copyable first API request, so those details must be established before deployment.

How do you use this agent?

After starting the backend, open /docs for Swagger documentation; after starting the frontend, use the interface to create or interact with agents in multi-turn conversations. Use the knowledge base for RAG questions, upload Swagger/OpenAPI definitions for custom tools, or run the HITL flow that generates an MCP server from OpenAPI information. The project says first startup checks for missing model configuration and presents guidance, but it does not document the exact configuration location, credential format, or first callable endpoint.

How does this agent compare with similar options?

FAQ