Agent Service Toolkit is a full-stack framework for building and deploying AI agent services using LangGraph, FastAPI, and Streamlit.
The toolkit addresses the challenge of moving from a prototype agent to a production-ready service by providing an integrated template that handles the complete pipeline. It bundles a LangGraph agent with a FastAPI backend that exposes both streaming and non-streaming endpoints, a Python client for service interaction, and a Streamlit frontend with chat capabilities. The architecture uses Pydantic for data validation and supports advanced features including human-in-the-loop interrupts, long-term memory via LangGraph's Store, and content moderation through Safeguard integration.
Teams building LangGraph agents should adopt this toolkit if they want to avoid scaffolding a service layer from scratch. It suits projects that need a chat interface quickly, support for multiple agents within a single service, and the ability to serve agents over both standard HTTP and the AG-UI protocol for compatibility with frontends like CopilotKit. The toolkit includes a RAG agent implementation using ChromaDB for teams interested in retrieval-augmented generation, and provides Docker support for containerized deployment. The Streamlit app offers voice input and output alongside text, with a sidebar showing previous conversations and a feedback mechanism integrated with LangSmith.
The project maintains active development with regular updates to incorporate new LangGraph features. The codebase includes comprehensive unit and integration tests covering the full stack. Documentation is provided for specialized components like the RAG assistant and AG-UI protocol support. The repository includes a working demo application accessible online, allowing prospective users to evaluate the interface and capabilities before adoption.
