Docs / Goals & loops

Goals & autonomous loops

Clioloop is built to not stop until the goal is reached. A standing goal turns a single request into an autonomous loop: after every turn a judge checks whether the goal is met, and if not, the agent takes the next step on its own.

Set a goal

/goal ship a working REST API with tests and a README

From there, each turn ends with a judgement. If the goal isn't satisfied, Clioloop continues automatically — across the terminal, the desktop app and gateway chats — with a live banner showing progress.

Control the loop

  • /goal status — show the active goal and progress
  • /goal pause · /goal resume — hold or continue the loop
  • /goal clear — drop the standing goal
  • /subgoal — add acceptance criteria the judge must satisfy

Acceptance criteria

Subgoals make "done" concrete. Add the conditions that matter and the judge will hold the loop open until they're all met:

/subgoal all tests pass
/subgoal endpoints documented in the README

Steer without stopping

While a loop runs you can nudge it without breaking flow:

  • /steer — inject a message after the next tool call
  • /queue (/q) — queue a prompt for the next turn
  • /background (/bg) — run a prompt in the background
Pair goals with Agentic Fusion for high-stakes work: the loop keeps going while a panel of models plans and reviews each draft. For parallel workstreams, hand tasks to the multi-agent Kanban board instead.
Goals & autonomous loops · Clioloop