ProPR
Log in Get ProPR

A second opinion on every pull request, without touching the code.

Comment /review on a pull request. ProPR checks the diff against the issue and the conversation, then posts numbered merge blockers, suggestions and a Score: N/10.

Critique loop Read-only
Review ContextIssue + prior changes + diff
AI ReviewerClaude/Codex
Score: 6/10 F1 merge blocker S1 suggestion
Review reads the initial issue, prior branch changes, and current diff, then returns structured critique without pushing code.

Product path overview

One model or several

Your default review model runs, or name models for several independent passes.

github.com/org/repo/pull/128/review astra fable51

Blockers, suggestions, score

Merge blockers (F1) name the required behavior, the evidence and the minimum fix. Suggestions (S1) are optional. A score and a copyable /fix line close the review.

Every finding gets an ID that never changes.

On public propr #2934, GPT-6 Astra blocked the merge on F1 and scored the branch 6/10. After the fix, the re-review found no blockers and scored 9/10.

A merge blocker states the rule, the evidence, and the minimum fix

A ProPR AI review by codex:gpt-6-astra on public pull request 2934 with an overall evaluation and merge blocker F1, To-do updates can attribute another owner's records to the caller, listing required behavior, evidence, and minimum fix.

Suggestions carry no obligation; the score closes the review

The same review with optional suggestion S1 and Score: 6/10.

The review ends with a ready /fix line

The review footer explaining /fix, /fix all, and named records such as /fix F3 S5, followed by a copyable /fix F1 S1 command.

After the fix, the re-review continues the numbering

The second ProPR AI review on pull request 2934 with no merge blockers, optional suggestions S2 and S3, and Score: 9/10.
Two review passes on public propr #2934.

Works on any pull request

Hand-written PRs included. Drop /review or /fix on any of them.

Focus instructions

Text below /review steers it. Ask for security, error handling, tests or a risky area of the diff.

Traceable results

Each review links to its task record, model, token usage and every later follow-up commit.

See the change, too

With visual previews on, the PR also carries captures of the changed screen. Public propr #2952 shows a before and after.

The reviewer sees past the diff.

A scout model gathers the unchanged callers, contracts, configuration and tests around the change and hands them to the reviewer, so findings rest on the whole codebase.

Review settings: GPT-6 Astra reviews pull requests, and a read-only Claude Opus 5.5 scout gathers related unchanged code first
The default review model, plus a scout model that gathers related unchanged code.

Scores roll up by model.

Analytics shows where each model starts, where review and fixes take it, and how many runs its pull requests need to merge.

Agent efficacy by model: evaluated PRs, initial and final review score, score delta, runs to merge and merge rate for Claude Opus 5.5, GPT-6 Astra and Claude Fable 5.1
Review scores roll up by model: where each agent starts, and where review and fixes take it.

Turn the findings you keep into commits.

Every review ends with a copyable line such as /fix F20 S3 S5. Paste it, trim the IDs you disagree with, and ProPR applies exactly those on the same branch. /fix alone applies the merge blockers; /fix all adds the suggestions. /ultrafix runs review and fix cycles toward your target score, within your cycle limit.

See how review becomes follow-up commits →