No more re-checking by hand
Every “done” is checked with a real build and the real tests it reaches. You read the verdict and its evidence, not the whole diff.
Time and money saved
Agents were meant to give you time back. If you still have to save time reviewing AI-generated code by re-running its tests and reading every diff, that time has only moved. Cotterpin does the checking, so you review a verdict with evidence instead of the whole change.
Good protection is invisible: the bug that never shipped looks like a quiet day. So Cotterpin shows its work. When a run ends, the session card shows what it checked, what it refused and what the run cost, and how much routing each stage to the right model saved. Every Monday, the weekly card adds it up: how many runs landed, how little time Cotterpin added, what you saved, and how quickly the plan pays for itself.
Coming in the next release These are examples with illustrative figures: the session and weekly cards ship in the next release. Your own cards will show your own runs, measured on your machine.
Cotterpin · session complete Task add rate limiting to /api/orders ✓ landed Duration 14m 32s Stages 6 Agents 3 Cotterpin fired 23× ✓ 19 ✗ 3 refused ? 1 asked Cost $1.84 API-equivalent Savings $1.21 saved (40%) vs all on Opus · high Tokens +66% extra tokens on a monthly plan ▸ 4 stages on Sonnet · medium, 2 on Haiku · low Details cotterpin runs show 7f3a
Cotterpin · your week · 29 Sep – 5 Oct Runs 64 58 landed 6 waited for you Agent time 15h 20m Time Cotterpin added 2m 41s (0.3% of agent time) Cotterpin fired 1,412× ✗ 87 refused, 5 destructive Cost $117.80 baseline $195.24 Saved $77.44 this week ▲ $9.10 at API prices Tokens +66% extra tokens on your plan (≈ $132 of a $200 plan) On track ≈ $335 / month · ≈ $4,000 / year Pro ($100/year) would pay for itself in under 2 days You ◆ Scout Bug Hunter grade 1 of 7 Plan Watch free (reports what Pro would enforce) Bugs 4 reported 3 verified 1 pending Rewards 14 points 1 free month earned Details cotterpin report weekly
Cost is the API-equivalent at published prices; savings compare Cotterpin's model and effort routing with running every stage on the tier your agent is set to use. You'll be able to hide the card at any time, and the full trail stays on your machine.
Every “done” is checked with a real build and the real tests it reaches. You read the verdict and its evidence, not the whole diff.
A false “done” goes back to your agent in the same session, with the failing evidence, instead of surfacing in review, in CI or in production.
Hand Cotterpin a task and get back a pull request with a reproducer test, the refuter's results and a receipt, ready to review.
Antigravity,Claude, Codex or Grok, under your own account. Cotterpin hooks in; you keep prompting the way you do now.
Install, run cotterpin init in your repository and cotterpin doctor to check your machine. The quickstart walks through it.
Every change is a commit you can revert. cotterpin uninstall removes it from a repository, and cotterpin --uninstall from your machine.
cotterpin report measures it on your machine, from the prices you set — nothing is uploaded.Put in your own figures. The starting values are examples, not measurements — replace them. Nothing you type leaves this page.
In development Bug Hunter is being built now and is not available yet. Below is the scheme we have decided on; details may change before it launches, and this page will change with them.
Find a real bug in Cotterpin and report it from your terminal. The report is built and redacted on your machine, and you see exactly what is sent before it goes. Every verified bug earns points, and points become free months.
Critical 5, major 3, minor 1. The first verified reporter of a bug earns its points. Confirming a known bug on another operating system or language pack earns 1.
On Pro or Team, points turn into free months: up to 6 a year. On the free Watch plan they build up as credit, applied when you subscribe. Credit lasts 12 months and is never paid out as cash.
From your second year, each consecutive annual renewal takes 10% off the next, up to 50% from year six. A lapse starts it again. Loyalty discount and free months both apply.
| Grade | Reached at (lifetime verified points) | Points per free month |
|---|---|---|
| Scout | 5 | 10 |
| Tracker | 25 | 10 |
| Bug Ninja | 60 | 9 |
| Bug Slayer | 120 | 8 |
| Predator | 250 | 7 |
| Exterminator | 500 | 6 |
| Bug Demon | 1,000 | 5 |
The higher your grade, the fewer points a free month costs, and the more you get: the beta channel, priority triage, a vote on the roadmap and a place in the Hall of Fame (opt-in, anonymous by default). For example, three major bugs and one minor make 10 points: one free month at Scout. Security bugs go privately to security@cotterpin.io. Your reports, their triage and the credit they earn will show on your developer dashboard, coming in the next release.
It routes each stage of the work to the model and effort it needs, instead of running everything on the strongest model, and measures what that saved against running every stage on the tier your agent is set to use.
Every week Cotterpin adds them up two ways: as extra tokens your monthly plan holds, and in dollars at published API prices. The figures are measured on your machine from your own runs.
Very little. Its checks run on your machine, and the weekly card shows how much time it added beside the agent's own time.
Watch is free, with no time limit, for as long as you like.