auth/callback.tshighOpen redirect risk
User-controlled `returnTo` is passed to `res.redirect()` without an allow-list. Validate against trusted origins.
Suggested: Add URL allow-list + normalize before redirect.
Best initial comparison
Native GitHub is unmatched for workflow and code hosting. MergeGuard adds an always-on AI layer: every PR opened gets a consistent, security-aware review with risk scoring and actionable fixes—without replacing how your team uses GitHub.
The same PR can look “fine” in chat—or carry hidden issues. MergeGuard turns the diff into structured signal before merge.
LGTM 👍
Fast human rubber-stamp. Easy to miss regressions when the diff is large or the reviewer is tired.
Risk score + findings are posted on the PR automatically—your team still owns the final merge decision.
GitHub is the platform. MergeGuard is the specialist reviewer that never skips a PR—and never forgets an edge case pattern you cared about last sprint.
Manual code review
AI-powered PR analysis
Automatic PR reviews
Security issue detection
Suggested code fixes
Risk scoring
PR summary generation
Large PR analysis
Manual
Team review consistency
Review in seconds
Detect hidden edge cases
Depends on reviewer
GitHub-native integration
Works automatically on PR open
AI-generated explanations
Multi-repo management
Manual
Usage dashboard
Smart review labels
AI fix recommendations
Developer onboarding simplicity
Manual setup
Real PR conversation patterns: structured review body, inline threads, and fix-oriented follow-ups—hosted entirely on GitHub’s UI.
Step 1
Pick repos and permissions. MergeGuard listens to PR events on your installation.
Step 2
Your team keeps the same branching model. MergeGuard wakes up when the PR opens or updates.
Step 3
Summary + inline comments + optional fix follow-ups—seconds after the latest push.
Teams ask hard questions before they let an AI near production code. Here is how MergeGuard is designed to respect your boundaries.
Repository-scoped access
GitHub App installation controls which repos MergeGuard can see—no broader org-wide read by default.
You choose the model path
Bring your own provider keys where supported; diffs are sent only for review jobs you trigger via PR activity.
Audit-friendly output
Findings live on the PR and in your workspace usage—useful for security reviews and post-incident retros.
Illustrative threads—your wording and severities vary by repo policy and model tier.
auth/callback.tshighUser-controlled `returnTo` is passed to `res.redirect()` without an allow-list. Validate against trusted origins.
Suggested: Add URL allow-list + normalize before redirect.
components/Form.tsxmediumSubmit handler sets `pending` but the button is not disabled—double submits can duplicate records.
Suggested: Disable button while `pending` and show inline spinner.
package.jsonlow`axios` was added but imports use `fetch` only—larger install surface without benefit.
Suggested: Remove unused dependency and refresh lockfile.
Measurable outcomes when every PR gets the same quality bar—not only the ones a senior had time to read line-by-line.
Seconds
Review latency
From PR open to first structured review on the diff.
Every PR
Consistency
Same checklist for interns and staff engineers alike.
Pre-merge
Early bugs
Catch issues while rollback is a revert—not a hotfix.
Humans
Focus time
Let people review architecture; let AI scan for foot-guns.
Install the app, open any pull request, and compare MergeGuard’s review to your usual pass—no workflow rewrite required.