Loop & Enhance

Loop & Enhance wraps a workflow run in a detect-fix-rerun cycle. After each round, Cerevisor looks at what failed, proposes edits to your workflow, applies them, and runs the next round.

The result: workflows that improve themselves between rounds, converging on a working version.

Plan requirement: included on the trial, Pro and Founding membership. Not on Free.

When to use it

• Your workflow mostly works but has reliability issues you can't pin down.
• You want Cerevisor to evolve the workflow toward something more robust.
• You're iterating on the shape of a workflow, not just its output.

When not to use it:

• For a workflow that already produces good output reliably: Loop & Enhance is overhead without payoff.
• For one-shot tasks where the cost of multiple rounds isn't justified.

Enabling it

Open the Run-mode dropdown (the chevron next to the Run button) and choose Loop & Enhance…. This opens the Loop & Enhance Setup Modal with the key settings:

Field	Default	Effect
Max rounds	3	Hard cap on cycles.
Budget cap (USD)	$1.00	Total cost across all rounds, including the planning between rounds.
Convergence criteria	Strict	Strict = stop only when there are zero failures of any kind. Lenient = also accept the result when the only remaining failures are intentional skips or fallback outcomes.
Pause between rounds	Off	If on, the run pauses after each round and asks you to approve continuing.
Planner model	Auto	The model that plans each round's edits. Auto follows the chat builder's model (falling back to your provider's default); or pick a specific model from your provider's list.

Your choices are saved with the workflow, so the next Loop & Enhance run starts from the same settings.

How it works

1. You click Run. The workflow runs round 1, same as a normal run.
2. When the round ends, Cerevisor sums up what went wrong: which agents failed, what their output looked like, and what the output checks said.
3. If nothing failed, the workflow has converged and the run stops.
4. Otherwise a planner model reads that summary and proposes edits to the workflow: the same kind of edits the chat builder makes, such as changing an agent's instructions or adding a skill.
5. The edits are applied to the workflow on your canvas.
6. Round 2 runs against the edited workflow.
7. This repeats until the workflow converges, or the round or budget cap is reached.

Safety rails

Several guards prevent a runaway:

• A hard cap on rounds, set per workflow. A run never exceeds it.
• A total budget cap covering both the rounds and the planning between them. The run halts when it is reached.
• Stuck detection. If a round ends with the same problems as the round before, the planner isn't making progress. Cerevisor raises a fixation warning and can stop the run.
• Limited edits. The planner can only make a restricted set of edits. It can't delete the workflow, change the provider, or do anything destructive.

The Loop & Enhance tab

When Loop & Enhance is running, the run window gains a Loop & Enhance tab showing:

• Round counter ("Round 2 of 3").
• Budget used so far.
• Per-round timeline:
• What failed.
• What the planner proposed.
• What was applied.
• The planner's reasoning.

Useful to understand why the workflow changed between rounds.

Rounds are also non-destructive: each round's workflow and outputs are saved as a version, and the Loop Version History modal lets you browse past rounds, compare them, and restore an earlier round's files back into your working folder.

Pause between rounds

If you turn this on, the run stops at the end of each round and waits for you to click Continue to next round. Useful when you want to review the proposed edits before committing.

The pause is between rounds, not between agents within a round.

Which provider plans the edits

The planner needs a model it can hold a conversation with, and Cerevisor picks one for it independently of the provider your agents run on. A workflow that runs on Codex or Cursor can still use Loop & Enhance: the agents run where they always run, while the planner works on a provider that can chat (typically your chat builder's or your default provider).

If none of your providers can do that, the run refuses to start with a clear error instead of failing mid-loop.

When convergence isn't possible

Some failures can't be fixed by editing the workflow:

• Provider rate limits.
• Network errors.
• Skill compatibility issues the planner doesn't know how to resolve.

In these cases, Loop & Enhance reaches the round or budget cap and stops: the last successful round's output is kept, and the summary explains what couldn't be fixed.

Cost expectations

Each round costs the workflow's normal run cost plus one planning call, usually a few cents on a small, fast model. A 3-round run of a $0.50 workflow with 3¢ of planning per round comes to about $1.60.

Set the budget cap to a value you're comfortable spending on an automated improvement cycle.

Why this exists

It's a different way of working: instead of "run, edit, run, edit" by hand, you set a goal and let Cerevisor drive the iteration. Worth it for stable workflows you keep coming back to; not worth it for one-shot tasks.