LIVE This is the project's own agent loop, compiled from src/ and running in this page. No API key and no backend: the loop drives the deterministic mock provider the repo ships for offline runs, so the planning is real and the prose is canned.

agent:~$ run --provider mock

Agentic AI

A ReAct agent: it reasons a step at a time, calls a tool when a tool helps, reads the observation, and stops when it can answer. Every step below is emitted by the loop as it runs.

Ask the agent

Run

Reasoning trace

• Nothing run yet.