These examples are complete applications, not isolated API fragments. Each one has its own Go module, immutable released dependencies, deterministic tests, an embedded skill, domain tools, a Harness Session, and a documented path to hosted or local models.

Choose an application

Application	What it teaches	Result
Weather Assistant	Skill loading, domain tools, streaming turns, events, and model replacement	A conversational weather recommendation
Research Assistant	A primary Loop, a delegated researcher Loop, retrieval, citations, and workspace artifacts	A cited Markdown brief

Both run without API credentials so you can inspect the runtime behavior first. The scripted inference clients implement the same inference.Client boundary as OpenAI, Anthropic, Google, Bedrock, or a local model adapter.