Delegate independent tasks

Enter a goal and choose an available persona. Each agent works in a separate local workspace, so you can keep tasks apart while they run.

Stopping a running task asks for confirmation and reports whether the request was accepted. Agents created from this panel run locally.

Inspect the plan and its progress

The Specs panel lists executable specs with their phase, status, and completed task count. Open one to move between its requirements, design, and task list.

This release supports browsing existing specs. Editing them inside the panel is not included.

Review the result in one panel

The session panel shows the latest completed response when available. A timeline brings the session's turns and log entries together in time order.

Review sessions link to their pull request and commit, so you can move from agent activity to the change it worked on.

See reuse readiness beside code health

Completed scans show a reuse readiness score from 0 to 100 on the scan card and in its detail view. Higher scores indicate code that may be easier to extract and reuse.

The score uses signals including dependencies, cycles, duplication, and complexity. It is an indicator, not a guarantee that code can be reused without changes.

Understand the change before the details

The summary describes the change and groups the affected files with their added and removed line counts. It can also include an assessment of alternative approaches.

Move from a finding to the next step

Findings include severity badges, supporting evidence, and guidance you can give an agent. A link opens the finding in the dashboard for follow-up.

Focus on the highest-priority standards

Filter standards to critical, high, medium, or low severity. Sorting the Severity column puts the most severe standards first, using the same order as review findings.