How the Agent Works

Once active, the agent follows the same three-stage loop for every finding Pulse AI flags.

From Finding to Confirmed Fix

1. Detect & Analyze

Pulse AI flags the sprint issue. The agent picks it up and revalidates it - live for a sprint finding, against the last synced state for a board one - rather than trusting a cached count. It then investigates the likely cause and impact.

2. Recommend

It explains what it found, which real issues are affected, and why it matters. For fixable findings, it asks whether you'd like it to prepare a fix - and stops there until you say yes.

3. Act with Confirmation

Only after you confirm does it show a full preview - one row per affected issue, what changes, and why - before writing anything. If you edit the plan, it shows the revised preview and asks again.

Nothing is written back to your sprint or board until you explicitly confirm it. There's no automatic or background editing.

Note: This loop plays out inside the Chat tab - see Chat →.