PR quality review
PURA can also review the pull request itself: title, description, labels and commit messages. Results appear in the PR overview comment, so every pull request meets your standard.
Features
AI code review for your whole org, on your models and your budgets.
Every plan covers all your developers and teams. You choose which model reviews what, and you decide how the budget is split across teams, authors and repos.
1-month free trial · bring your own keys · we never store your code
Pricing
Unlimited users and teams at a fixed price
Per-user tools charge a fee for every developer. PURA charges by the number of reviews, so adding people, repos or teams doesn't change your bill.
- Unlimited contributors, repositories and teams on every plan, including the trial.
- You pay for reviews: one review is one pull request PURA reviews. Drafts and pull requests no agent picks up are never counted.
- No overage charges. When you reach your monthly allowance, reviews pause until the 1st of the next month, or you move up a plan.
- Model usage is billed by your own AI provider at their price. PURA adds no markup.
Saved per year
$10,000
For 50 developers and 1,000 reviews a month: CodeRabbit Pro at $24 per developer a month, against the PURA Pro plan plus about $0.20 of model usage per review. Annual list prices.
Build or buy
Cheaper than running it yourself
A home-grown reviewer needs a database, servers to run the reviews, and an engineer to keep it working. PURA costs less than that upkeep, and it is ready today.
What you get without writing or maintaining any of it:
- Reviews posted inline on GitHub, with suggested fixes you can apply in one click.
- Anthropic, OpenAI and Google models, plus OpenAI-compatible providers at the token prices you enter.
- Budgets checked before every review, with a fallback to your next provider.
- Spend analytics by team, author, repo, model and provider.
- Secrets redacted before your code reaches a model.
Saved per year
$4,600
Building your own: $50 database, $100 compute and 4 engineer hours at $100 an hour each month, against the PURA plan for 1,000 reviews a month. Model usage costs the same either way, so it is left out.
Model control
You control which model reviews what
With a black-box tool, someone else picks the model and you live with the results. With PURA, you choose the model for every kind of review and can change it whenever you like.
- org all repos
- your default model
- team payments
- a stronger model
- use case auth changes
- your strongest model
- use case docs only
- skipped
- Org-wide: set a default model for every review agent.
- Per team or repo: give a team its own agent with its own models.
- Per use case: each agent has a default, high and low model, and a plain-English rule picks one for each pull request, or skips it.
- Security first: keep your strongest model on auth, payments and other sensitive code.
- A/B test: run a new model for one team or a few repos, compare findings and cost per model in analytics, then roll it out.
- Your keys: connect your Anthropic, OpenAI, Google or OpenAI-compatible key. Keys are encrypted and never shown again after you save them.
---
name: security-review
description: Security review for auth and payments code.
.pura-routing: |
Use the high model for changes under auth/ or payments/.
Use the low model for small refactors.
Skip pull requests that only change docs.
---
Check for injection, broken access control and unsafe
handling of secrets. Ignore formatting.Budgets
Budgets split by team, author and repo
A single org-wide limit can't tell a busy payments team from a quiet docs repo. PURA spreads the budget the way your org is organised, and enforces all of it from one API key.
- contributing team payments
- $400 / month
- reviewing team platform
- $250 / month
- author @dana
- $40 / month
- repo checkout-svc
- $20 / day
- each pull request labelled release
- $5 / week
Example budgets. You set your own amounts.
- Budgets per contributing team, reviewing team, author, repo or single pull request.
- An overall budget per agent, and a separate budget per AI provider.
- Daily, weekly and monthly limits, in any combination.
- Allow and block lists decide which teams, authors and repos each agent serves.
- When a provider reaches its budget, PURA moves to your next provider. When a budget runs out, PURA pauses those reviews and says which budget was reached.
And more
More than code review
Review-agent skill
Each review agent follows its own SKILL.md in your repo: what to look for, what to ignore, and which model to use. It is read from your default branch, so a pull request cannot change its own reviewer.
Custom skills
Give reviewers your own skills, such as a security checklist or your API conventions, shared from one repository across every agent. Included: 10 on Scale, 25 on Pro.
In every review
One clear review per pull request
However many agents review a pull request, your team gets one review in GitHub.
- Runs when a pull request is opened or updated, or on demand when you comment
/pura review. - A summary, findings ranked P1 (must fix), P2 (should fix) and P3 (consider), and suggested fixes you can apply in GitHub.
- Clean pull requests are approved. A must-fix finding requests changes.
- Follow-up reviews resolve fixed findings and don’t repeat old ones.
- Reply /pura ignore to dismiss a finding on this pull request, or /pura ignore always to stop it coming back.
- Secrets are redacted before code reaches the model, and PURA doesn’t keep your code after the review.
Get started
Try PURA on your own pull requests
Install the GitHub App and your first review runs on the next pull request. The trial lasts a month and needs no card.