Hands-on OpenAgent Tutorials
Build a self-hosted AI assistant, answer questions from your documents, and connect an MCP order lookup tool with working examples and screenshots.
Build something you can verify
Each tutorial includes the setup, sample inputs, expected results, real OpenAgent screenshots, and troubleshooting. Start with a chat assistant, then reuse that model connection for document retrieval or MCP tools.
Self-host an AI assistant
Run OpenAgent locally, connect a model, and verify your first answer and conversation history.
Chat with an employee handbook
Download a sample handbook, index it, and check both cited answers and questions it cannot answer.
Connect an MCP order lookup
Run a small Python MCP server and inspect a real tool call, arguments, and shipping result.
The screenshots use a local Windows instance built from OpenAgent source on October 31, 2026. The model is Qwen through an OpenAI-compatible endpoint; the document example uses Alibaba Cloud text-embedding-v3. You can use another supported model with the capabilities each tutorial requires. Generated wording may differ; use the listed checks to verify the result.
All employee policies and orders in these examples are fictional. Self-hosting OpenAgent keeps the application and its database on your machine; a cloud model or embedding service still receives the text sent to it. Choose local providers when the whole inference pipeline must stay local.