Remediation

Context describes what Polylane sees. Detection describes what it spots. Investigation describes how agents work out what happened. Remediation is where agents end the incident: reversible operations run where you allow them, and the permanent fix follows as a pull request for your review.

The surfaces

• Autofix: when an agent has a clear fix, it pushes a branch and opens a pull request in your repository. Autofixes start from an automation, a chat thread, or an investigation, and run on the native engine or an external coding agent such as Devin, Cursor, or Factory.
• Automations: agents running on their own. Pick a trigger such as polylane.change_record.created, pick the actions the agent may take, and pick the destinations where results land. Templates cover the common setups.
• Skills: reusable, named agent workflows. Package a remediation you keep running by hand so threads, automations, and other agents invoke it the same way every time.

The autonomy dial

You decide how far agents go. The dial has three settings, and each one is enforced in the platform, not left to the agent's judgement.

1. Read-only by default. New cloud accounts connect in read-only mode. Agents can query anything but every write call is refused until a workspace admin turns off read-only in the account's settings.
2. Reversible actions where policy allows. On accounts with writes enabled, agents run reversible operations such as restarting a service or rolling back a deploy. Every write call still passes a safety review before it executes, and writes are only available in chat threads, never in background runs.
3. Code always gated by your review. Agents never push to your default branch. The permanent fix arrives as a pull request, so your continuous integration checks run and a human merges it. Polylane records who merged each autofix.

When to reach for which