Give the agent a goal (todo queue — 24/7 daemon works through it)
"act" tasks run on the host through the real skills (terminal · browser · memory) in an auto-continue loop until the goal is done — results + evidence land below. Queued goals survive restarts (disk-backed queue) and are picked up by the 24/7 daemon every cycle (~60s).
Pending
loading…
Done (newest first)
loading…
Daemon journal (live)
…
agent@host:~$
Teach it something permanent
Facts here are recalled into every relevant chat automatically (keyword match) — the AI never forgets. Chat history is also stored permanently on disk (logs/chat-history.jsonl).
Facts (0)
loading…
Recall test
Activity journal
…
Skill roster
queued "act" goals combine these skills autonomously.
loading…