The AI Agent Service Toolkit is a complete, open-source template for running an AI agent as a production-style service. It combines a LangGraph agent, a FastAPI backend, a reusable Python client and a Streamlit chat interface into one coherent project you can fork and adapt.
Many agent demos stop at a notebook; this toolkit covers the whole path from agent definition to a served application with a user interface. The LangGraph agent implements current framework features including human-in-the-loop interrupts, flow control with commands, long-term memory via the Store, and supervisor-style multi-agent setups. The FastAPI service exposes both streaming and non-streaming endpoints, with a streaming design that supports token-based and message-based updates, and every agent is also served over the AG-UI protocol for compatible frontends such as CopilotKit. Multiple agents can run in one service and be addressed by URL path, a RAG agent backed by ChromaDB is included, and the Streamlit app adds chat history, voice input and output, and a feedback mechanism wired to LangSmith. Docker Compose files, a private-credentials pattern, and a solid unit and integration test suite are included. The project is MIT-licensed, has more than 4,500 stars on GitHub, and was under active development in 2026.
Free. The toolkit is open source under the MIT license; you supply your own LLM API keys and hosting.
Developers starting a LangGraph project who want a proven, end-to-end starting point โ agent, API, client and UI โ instead of assembling service plumbing themselves, and teams prototyping internal agent services.