Approve. Comment. Vetoo.
Agentic AI code review for GitHub and GitLab. Tailored to your team.
import { db } from '../db'import { signToken } from './jwt'export async function refreshSession(token: string) { if (!token) throw new AuthError('missing token') const user = await db.query(`SELECT * FROM users WHERE token=${token}`) const user = await db.query('SELECT * FROM users WHERE token = $1', [token])String-interpolated SQL on token is injectable. A crafted token could read any row. Bind the value instead.
@vetoo-review is this a real risk or theoretical?
Real. This is the refresh path, so it is reachable without a valid session. Nothing upstream filters the token before line 41.
const next = signToken(user.id) await cache.set(`session:${user.id}`, next)The rotated token is cached without a TTL. Callers assume a 15 minute expiry.
return next}7c2e9b1 into mainYou're all set. The feat/token-refresh branch can be safely deleted.
- GitHub
- GitLab
- Linear
- Notion
- Slack
- Jira soon
Top 4 in precision.
Independent Martian benchmark of 22 AI code reviewers on real, human-reviewed pull requests.
How it works
It reads beyond the diff.
Integrations and your other repositories.
Multiple agents review your code from different angles.
Signals merge and rank. The stricter call always wins.
One verdict lands.
Severity, reason and inline comments with fix proposals.
Your code, your rules.
Vetoo leaves comments.
Give feedback by reacting to them.
Feedback collected and evaluated.
Learnings applied as rules.
Predefined rules in plain Markdown.
## Review rules- Block any PR that logs PII- Treat missing tests as nits- Blocking
logger.info prints the user email
login.ts:27 - Nit
New refresh path has no test
session.ts - Nit
Prefer early return over nested if
session.ts
Code review, not autopilot.
- One model behind a proxy with an AI review sticker
- A coding agent let loose in your repo to write and ship
- Review bolted onto a do-everything platform
- Your source kept around to train a model
- A fresh review the moment a pull request opens
- Ranked findings, tailored to your team
- One job done best, with a mix of models, including frontier ones
- Your code deleted the moment the review ends
Code reviewed, then gone.
Kept
- Pass, comment and block verdicts
- Your votes as review preferences
- Suppression rules you set
- Repo and pull request identifiers
Never stored
- Your source code
- Secrets, tokens or env vars
- File contents after the review
- Anything used to train a model
Start free. Scale by the seat.
In every plan
- Inline review comments
- Review from every angle
- Memory: learning from usage and feedback
- Cross-repository checks
- Interactive comments
- Context enrichment with Notion, Linear and Slack
- Review effort levels
- PR checks
- Unlimited repositories
1 credit = $0.01 Reviews from $0.70 Credits last 12 months
Resources
Compare
Vetoo vs CodeRabbit
Focused review quality over platform complexity.
ComparisonVetoo vs Greptile
Multiple agents on models from several providers, across every repo a change touches, with none of that data retained.
ComparisonVetoo vs Bugbot
Independent review without ecosystem dependence or vendor lock-in.
Learn
What is AI code review?
How agentic review reads a pull request, gathers context, reaches a verdict, and fits alongside a human reviewer.
ChecklistThe code review checklist
A practical pass covering correctness, security, tests, readability, and design.
FAQFrequently asked questions
Multi-agent review, data handling, GitLab support, pricing, and team memory.
FAQ
How is Vetoo different from other AI code reviewers?
It adapts to your team, and it stays quiet unless it has something worth saying. Every reaction on a finding becomes part of your team's memory, alongside the rules you write in VETOO.md, so each review gets closer to how your team works. Multiple agents review every change from different angles and cross-check each finding before it posts, so what reaches your pull request is curated and verified: fewer comments, more of them worth acting on.
What is agentic AI code review?
Agentic AI code review uses autonomous agents that read a pull request, gather context and act on it, instead of a single model pass over the diff. Vetoo runs multiple agents in parallel, each reviewing the change from a different angle, and only a curated set of findings reaches your pull request.
Which AI models does Vetoo use?
Vetoo runs agents on a mix of models, including frontier ones, side by side, so no single model decides the review alone. Each one surfaces a different slice of issues, and only a curated set of findings reaches your pull request, not the verbose output of a single model.
Does Vetoo store my source code?
No. Vetoo only holds your code while the review runs, encrypted in transit and handled in isolation, and it is gone once the verdict lands. Review comments and their metadata are kept for a limited time so conversations and team memory keep working.
Can I ask Vetoo questions about its review?
Yes. Reply to any finding, or mention @vetoo-review in a comment, and it answers in the thread. Ask why something was flagged, push back on a finding you disagree with, or ask for a fresh review pass.
What programming languages does Vetoo support?
Vetoo is language agnostic. It reviews any language the underlying models understand, which covers every mainstream language and most niche ones. VETOO.md rules, conversations and team memory work the same in every language.
Can I self-host Vetoo?
Not today. Vetoo is a hosted service. Each review runs in an isolated sandbox, and your code is deleted once the verdict posts. On Enterprise you can bring your own model key, so reviews run on your own provider account.
Does it work with GitLab?
Yes, and all features are available.
Can we customise the review?
Yes. Commit a VETOO.md file to set severities, tone and focus. Votes on findings tune it further.
Clear the review queue before it forms.
$20 in credits on sign-up.