AI product 2025 — today
AI writes the code. Pullminder checks it.
A verification step for teams that code with Copilot, Cursor and Claude Code. We designed, built and launched it ourselves.
- <60s to scan a pull request
- 210+ detection rules in 25 rule packs
- 0 lines of customer code stored
The problem.
Teams now write code much faster with Copilot, Cursor and Claude Code. Review didn’t speed up with them: the same people face far more pull requests, and some of the code reads as correct while it hides an SQL injection, broken login logic or missing tests.
Pullminder is our answer to that, and our own product from the first market analysis to the launch.
What we built.
A check that runs between the AI editor and the production branch, before a person reviews anything.
- A risk score from 0 to 100 for every pull request, from its complexity, its security impact and how likely it is to be AI-written.
- Security scanning on the diff, with 25 rule packs and 210+ rules.
- Merge policies set in a visual wizard or in YAML, which block a pull request that fails them.
- A brief in plain English for every flagged pull request, so the reviewer knows where to look.
From analysis to launch.
In late 2025 we studied the market: CodeRabbit, Graphite and SonarQube, and the gap between them. In the first quarter of 2026 we built the risk engine, the scanning, the policies and the CLI; in the second, the brand, the design system, the website, outcome-based pricing and a 90-day go-to-market plan.
In the third quarter the Founding 30 programme opened: half price on the Growth plan for the first 30 customers.
