Task board
Koja Agent tasks moving through their review-gate lifecycleLoading tasks…
Validator panel
Cheap + expensive routes per task — the Step-4 augmented checkpoint- Loading…
Delegation & learning
Which models the agent delegated to, AI cost, and how often a human was neededDelegation by model
Validation route
Calendar
Scheduled jobs — the scheduler wakes your agent machine when due, then suspends it idle- Loading…
Learning loop
The agent's mission, and the feedback that shapes how it worksMission — one line
The agent's top priority and boundaries in a single line — it leads the prompt on every task. Write it in your own voice: "I", "me", "my" mean you (the agent is told first-person refers to you, not itself). Anything you paste is compressed to one line.
SOUL — the agent's character
The longer description of who your agent is and how it works — its persona, standing instructions and judgement. This is its system prompt (the Hermes SOUL.md); the one-line mission sits on top of it.
Feedback — minor notes auto-apply, major ones are proposed for review
- Loading…
Memory log
Chat, files & web-search history — the lighter record (full review gates apply to agent tasks only)- Loading…