Get started¶
The first AgentMuru run is deliberately local and deterministic. Start with FakeModel to
learn the runtime and Workspace, then switch the generated project to one official provider.
Recommended path¶
- Install AgentMuru and run the environment checks.
- Complete the quickstart with a credential-free model.
- Tour the Workspace and learn what each panel means.
- Use a real model from OpenAI, Anthropic, or Google.
- Govern a tool before connecting side effects.
Choose another entry point¶
| Goal | Start here |
|---|---|
| Add AgentMuru to an existing Python project | Installation |
| Create a durable local application | SQLite persistence |
| Implement a custom model adapter | Provider contract |
| Deploy the HTTP and WebSocket server | Deployment |
| Explore the Go-native experiments | Labs |
What you will have¶
At the end of the recommended path you will have a Python module exporting one
Application, a browser Workspace at http://127.0.0.1:8000, and a model provider that can
stream text and tool calls into the same runtime contract.