Start with intent
Open a plan from a prompt, repository improvement, GitHub issue label, todo, or follow-up request.
ProPR carries each change through planning, an isolated run, automated review, and PR refinement. The result arrives in GitHub with the decisions, commits, and follow-up history attached.
Open a plan from a prompt, repository improvement, GitHub issue label, todo, or follow-up request.
Choose the repository, branch, files, screenshots, task size, and model before code changes, using the context preview to see what the planning model receives.
Refine generated tasks until the plan is clear, then approve one focused change or split it into smaller pieces.
The agent gets a separate Docker workspace and Git worktree, leaving the main checkout untouched.
Run a read-only review pass that posts an overall evaluation, severity-grouped findings, and a score on the PR.
Use normal comments and review comments for ordinary follow-up work, /review for AI critique, /fix to apply pending AI review suggestions, and /merge, /switch, or /ultrafix for branch, model, and review-fix controls.
ProPR keeps the important decisions visible: what you asked for, which plan you approved, which agent ran, what changed, what /review found, how feedback landed, and what entered GitHub. For concrete end-to-end walkthroughs of this path, see the recipes.
ProPR shapes the task into a plan — selectable context, refinement, and approval — before any code changes.
Commits, logs, reports, file changes, cost, and PR status stay attached to the task.
Pull request comments and review comments become part of the agent loop, from PR fine-tuning follow-up commits to the actions available through GitHub comment control.