Open-source AI code review

Review pull requests with AI that thinks like an experienced engineer.

PRAuditor runs a multi-agent review over every pull request — security, performance, logic, and readability — and posts clear, structured feedback before you merge.

Add checkout flow#128
PRAuditor
4 findings1 critical2 major1 minor
Securityauth/session.py:42

Session token compared with `==`; use a constant-time compare to avoid timing attacks.

Performanceapi/orders.py:88

Query runs inside the loop — an N+1. Fetch all orders in one query before iterating.

Logicutils/dates.py:15

Off-by-one on the range end — the last day is excluded from the report.

Readabilitycomponents/Table.tsx:120

Deeply nested ternary; extract to a small helper for clarity.

Review posted to the pull request

The workflow

What using PRAuditor actually feels like

The whole journey lives inside the GitHub flow you already use — no new tools, no context switching.

Step 1

Connect your repositories

Install the GitHub App on the repos you care about — two clicks, no config files, nothing to wire into CI.

Step 2

Open a pull request

Keep working exactly how you do today. The moment a PR opens or gets new commits, PRAuditor quietly wakes up.

Step 3

Get a real review

It reads every changed line and leaves clear, categorized feedback right on the PR — security, logic, performance, and readability.

Step 4

Merge with confidence

No more rubber-stamping code nobody actually read. Ship knowing it was genuinely reviewed.

Features

Everything a thorough reviewer checks

Four review agents, plus the workflow and history around them.

Security Analysis

Flags injection, unsafe deserialization, hard-coded secrets, and missing auth checks before they ship.

Performance Review

Catches N+1 queries, needless loops, and expensive work in hot paths on changed code.

Logic Detection

Surfaces incorrect conditions, missing edge cases, and regressions a tired reviewer would miss.

Readability Improvements

Suggests clearer names, less nesting, and comments where the logic actually gets hard.

Structured Markdown Reviews

Posts a clean, grouped review comment back to the PR — by file, severity, and category.

GitHub Integration

Installs as a GitHub App. Reviews every pull request automatically via webhooks.

Review History

Every review is stored, so you can revisit findings across a repository over time.

Coming soon

Risk Scoring

A single score per PR from severity mix, complexity, and file count.

Live example

See a review, end to end

Changed code, the AI's findings, and a suggested fix — no backend required.

auth/session.py
def verify_token(token, expected):
- if token == expected:
+ if hmac.compare_digest(token, expected):
return True
return False

Token compared with `==` — vulnerable to timing attacks. Use a constant-time comparison.

Suggested fix
-if token == expected:
+if hmac.compare_digest(token, expected):

Ready to improve every pull request?

Install the GitHub App and get your first AI review in minutes.