Review is the bottleneck. Price it properly.
AI writes the code faster than your team can review it. Here is what that queue costs you every month, what PURA costs to clear it, and where the difference lands.
Engineering hours returned
0 hrs/mo
The first pass PURA takes off your reviewers, every month, at the reference volume below.
1,000 reviews × 30 min of reviewer time each × 30% first pass ÷ 60 = 150 hrs
Every year
1,800 hrs
About 1.1 full-time engineers' worth of first-pass reading.
That time, in money
$15,000
$180,000 a year at $100/hr loaded.
Net of what PURA costs
$14,633
After the $367/mo bill, every month.
Your agents write faster than your team can review
Coding assistants multiplied how much code gets proposed. Review capacity is still one human at a time — so the queue, not the keyboard, sets your delivery speed.
The shaded wedge is the backlog: work that is finished but not merged. It is the most expensive inventory an engineering team holds — paid for, and earning nothing until it ships.
What the queue costs you every month
At 1,000 pull requests a month, the first pass alone consumes 500 reviewer hours — reading diffs for the mechanical things before anyone gets to the design questions that actually need a human.
PURA takes that first pass. Findings are waiting in the PR before a reviewer opens it, so the human starts from a summary instead of a blank diff — and the author gets feedback in minutes rather than tomorrow.
First pass, today
500 hrs
Reviewer hours consumed every month
Taken off the team
150 hrs
30% of that first pass
That time, in money
$15,000
$180,000 a year
Net of what PURA costs
$14,633
After the $367/mo bill
Waiting for a human
PURA's first pass
And the side of the ledger you pay
Against those returned hours, here is the whole cost of running it.
Per review
$0.00
All in — platform fee and model spend together.
Per developer
$0.00
And it falls every time you hire.
Surprise invoices
0
Budgets pause reviews instead of overspending.
Reference team: 50 developers, 1,000 reviews a month, PURA Pro.
Two line items. That’s the whole bill.
No seats, no overage, no support tier, no per-repo add-on.
PURA platform fee
$0/mo
Paid to us. Fixed for the month, whatever your traffic does.
Model spend
$0/mo
Paid to your own provider, at cost. PURA never marks it up.
All in
$0.37 per review · $7.33 per developer
$0/mo
Routing is the lever
A dependency bump does not need a flagship model. Sending routine work to lean models is where most of the model spend disappears.
Everything premium
$617
$0.45 a review
With routing
$367
$0.20 a review
Saved per year
$0
41% off the no-routing bill
Hiring doesn’t move the number
Contributors, repositories and teams are unlimited on every plan. Only review volume changes the price.
Developers
50 → 60
Ten more engineers this quarter.
Added to your bill
$0
There is no seat to buy.
Cost per developer
$6.11
Down from $7.33.
The worst month is a number you choose
Usage-based AI tooling usually means the bad month is a surprise. PURA inverts that: you set daily, weekly and monthly caps per agent, team, repository and provider, and when one is reached reviews pause instead of spending more.
There is no overage line on the invoice, because there is no overage. The allowance resets on the first of the month.
How budgets workSpend against your cap
Illustrative month. Once a cap is reached PURA stops routing reviews to that agent, team, repo or provider until the window resets.
What you get for it
One lever you can put a number on today, and four that keep working every month after.
The right model per PR
Routing rules send docs, typo fixes and dependency bumps to lean models and keep premium models for the code that matters.
$3,000
Model spend at cost
Your own OpenAI, Anthropic or Gemini keys. Your provider bills you directly — no reseller margin on inference.
No markup
Growth is free
Unlimited contributors, repositories and teams on every plan. Hiring never lands on the review budget.
$0 / dev
Budgets that actually stop
Daily, weekly and monthly caps per agent, team, repo and provider. Hit one and PURA pauses rather than overspending.
No overage
Spend you can attribute
Cost broken down by repo, team, user, model and provider — visible before the invoice, not after.
Full visibility
How these numbers are built
- Every figure on this page describes one reference team: 50 developers running 1,000 reviews a month on the PURA Pro plan. Your own numbers will differ.
- The platform fee is the plan whose monthly review allowance covers that volume — $39, $99, $199 a month across the self-serve plans. Annual-commitment list price, shown amortised across 12 months.
- Model spend is added on top, because PURA is bring-your-own-key: your provider bills you at cost and PURA takes no cut. We assume $0.20 a review — a mixed-routing blend for a review of roughly 60k input and 6k output tokens.
- The "everything premium" baseline assumes every pull request goes to a premium model instead, at $0.45 a review. It compares PURA against PURA without routing rules, not against any other product.
- Engineering time is valued at 30 minutes of reviewer attention per pull request, of which 30% is first-pass work an AI review absorbs, at $100 an hour fully loaded. These are planning assumptions, not measurements — halve them if your team reviews faster, and the case still holds.
- PURA does not replace human review. Its findings are advisory and a person still approves every merge, so the time figures describe a first pass your reviewers no longer start from scratch — not headcount you can remove.
Illustrative, not a quote. Your real bill depends on the models you pick, how well your routing rules match your codebase, and how much of your PR traffic you route through review at all.
Questions about cost
What finance asks before signing off.
What does PURA actually cost?+
A flat platform fee for your plan, plus the model spend you pay your own provider directly. There are no per-seat fees, no overage charges, and no markup on inference.
Why is model spend not included in the plan price?+
Because PURA is bring-your-own-key. You connect your own OpenAI, Anthropic, or Google account and that provider bills you at cost. Bundling it would mean marking it up, and it would hide the one number routing rules and budget caps are designed to bring down.
What counts as a review?+
One pull request that PURA reviews and routes to one of your agents. Draft pull requests never trigger a review, and PRs that select no review agent — no-match or budget-exhausted — are never counted. Allowances reset on the first of each month.
What happens if we go over our review allowance?+
PURA pauses reviews for the rest of the month rather than charging overage. You can move up a plan at any time, and per-agent, per-team, per-repo, and per-provider budget caps let you decide which work keeps running as you approach the limit.
How do I lower my model spend per review?+
Write routing rules that match the risk of the change. Send docs, typo fixes, and dependency bumps to a lean model and reserve premium models for auth, payments, and core logic. The spend analytics dashboard shows which repos and teams drive cost so you can tune the rules that matter.
Do we pay more as the team grows?+
No. Contributors, repositories, and teams are unlimited on every plan — the only thing that moves the price is review volume. Adding twenty engineers who each open a couple of PRs a month may not change your plan at all.
Run it on a real pull request.
Install the GitHub App, connect a key, and see the first review in a couple of minutes. One month free — no card, no seat count, no sales call.