Agent Service Toolkit

Agent Service Toolkit is a Python template for developers building their own AI agent services. It brings together agent logic, an API and a browser chat interface so you don't have to build each part from scratch.

• Custom agents: Build LangGraph agents with human approval steps, long-term memory and supervisor coordination, and serve multiple agents through one service.
• Chat interface: The Streamlit app supports voice input and output and lets users return to previous conversations for each agent.
• API and app connections: FastAPI serves complete responses or streams tokens and messages, while the included client and AG-UI support connect agents to other apps, including CopilotKit.
• Document retrieval: A basic RAG agent uses ChromaDB to retrieve information for answers.
• Moderation and feedback: Safeguard content moderation requires a Groq API key, and conversation ratings integrate with LangSmith.

The toolkit is open source under the MIT license and runs locally or on your own server using Python or Docker. It supports local LLM backends through Ollama, vLLM and SGLang, as well as cloud providers including OpenAI and VertexAI. The service and chat interface can run on your hardware; choosing a cloud model sends model requests to that provider and requires its credentials.