Background tasks

Agents can run long-lived work in the background — a build, a test watch, a training run — and (optionally) attach a monitor that wakes the model with new output on an interval. The CLI lets you see and stop these tasks from the host.

List

okto tasks list                  # aggregate across lair + all known agents
okto tasks list --agent lair-myrepo   # just one agent

Columns: task id, agent, status (running / done / error / cancelled), started, and the command (first line, truncated). Task state is read from disk:

• lair: ~/.okto/lair/session/tasks.json
• agent: ~/.okto/agents/<name>/data/session/tasks.json

Stop

okto tasks stop <id>                    # a lair-local task
okto tasks stop --agent lair-myrepo <id>   # an agent-local task

The response reports fired: true if the task was actually running and has now been cancelled, or fired: false if it was already finished or the id doesn't exist.

Tasks vs. notifications

A monitored background task can wake you with push notifications when it produces output or completes — handy for long runs you don't want to babysit in the chat.