What's inside

A small set of parts that fit together into one dependable agent — nothing you can't reason about.

A conversation loop that plans, calls tools, observes, and keeps going until the task is actually finished.

One process that dispatches the same agent across every chat platform at once, with shared memory.

Procedural memory — it turns a solved problem into a reusable skill it can call again later.

40+ tools with swappable backends — run them locally, in Docker, or in the cloud.

Agent-curated memory plus context files that shape every conversation the agent has.

A built-in cron engine and parallel subagents for isolated, long-running work.