# Autopilot

Autopilot schedules supported project workflows according to the project’s effective policy. It does not bypass eligibility, integration readiness, active-experiment, cooldown, or capacity checks.

```serpclimber-doc
{
  "type": "projects-autopilot-policy"
}
```

## Policy behavior

When **Draft new experiments automatically** is enabled, Autopilot selects Opportunities that meet the project’s Pages criteria and prepares experiment Drafts. Candidate selection is an internal step, not a separate setting or visible Page state.

Autopilot can also publish Drafts, collect evidence, and apply due decisions according to the configured policy. Manual policies stop at their configured review points. Automated publishing or decisions can create external consequences, so review the policy preview before enabling them.

## Effective settings

Project overrides take precedence over workspace defaults. Pages criteria, cooldowns, and workspace concurrency belong in Project or Workspace Settings rather than being duplicated in Autopilot. Drafting models are managed by SERPclimber.

Changing Autopilot affects future scheduler behavior. It does not silently rewrite a running experiment. A settings confirmation explains which future work may be enqueued, suppressed, published, or decided.

## Monitoring

Autopilot shows jobs, failures, capacity, and recent activity for the project. Routine jobs can be retried after their blocker is fixed. Use Chat for a prioritized operating brief that combines failures, Drafts, running Experiments, due decisions, Opportunities, and current capacity.
