About Rerunner
Rerunner is an independent, early-stage company. We're building regression tests for AI coding agents: CI for the files and settings that decide how an agent behaves.
Why we're building this
Teams now change their agents' instructions as often as their code, and review those changes by reading them. Reading doesn't tell you how an agent will behave. Running it does.
We want a change to CLAUDE.md to arrive with the same kind of evidence before merge that a change to source code already gets. The essay sets out the argument in full.
Principles
How we make decisions when they're hard.
- Evidence over scores
We report what we measured, task by task, with counts. No composite score, and no percentage we haven't calibrated.
- Honest verdicts
When a difference could be noise, we call it inconclusive. A gate that blocks on noise gets switched off, so we only block on facts.
- Vendor-neutral
We don't make a coding agent. We start with Claude Code and intend to test whichever agents your team uses.
- Security first
Running agents on pull requests means running untrusted code. We design for that from the start, and say plainly what we can't promise.
- Nothing made up
We have no customers yet, so you won't find logos, testimonials or usage numbers here. Sample reports are labelled as illustrations.
Status
In development · early access. Rerunner starts with Claude Code; Codex, Cursor, Gemini CLI and OpenCode are on the roadmap. We're pre-revenue and looking for design partners.
Rerunner is not affiliated with or endorsed by Anthropic, or by any other agent vendor.
Team
Rerunner was founded by Seçkin Büyük and is built by a team of four.
Contact
General questions and early access: hello@rerunner.dev.
Security reports: security@rerunner.dev.