Use case
Ship AI written code only with evidence for every criterion
Short answer
Reqursor Development links every acceptance criterion to the lines of the diff that satisfy it, runs your own checks as hard gates, and has a second agent verify the change before a security scan and the ship decision.
Every criterion points at lines
Each ticket carries an acceptance criteria checklist. The reviewer returns a structured verdict that cites file and line evidence for every criterion, and the run view links each criterion to the exact lines in the diff.
No evidence, no ship. A ticket that cannot be checked off does not ship silently wrong.
Your own checks decide, not the model
Your build, typecheck, lint and tests run as hard gates, together with built-in scans for secrets and stubs and a dependency policy. Checks use no AI, they pass or fail, and one failing check vetoes the ship whatever the reviewer concluded.
Diff coverage and a runtime smoke test are optional, and any command you can run in a terminal can become a check. A missing tool, such as a compiler that is not installed, is reported as an environment problem for you to fix, never patched around by the agent.
A second agent that did not write the code
The verifier runs in a fresh session and sees only the diff and the ticket, never the implementer’s reasoning. It works out again whether the work is done.
A security scan follows. Low risk findings are fixed automatically and checked again, and failing checks or review findings get a limited number of fix attempts, each one checked again from scratch.
The ship decision and the record it leaves
A change ships only if the verdict says ship, every required criterion has evidence, every check is green and no budget was breached. Anything short of that lands in Needs review with an escalation report, instead of a half-finished commit.
Large or sensitive diffs wait for a person to approve them. An append-only audit log records every run, decision and settings change, and you can download the full trace of any run as Markdown or JSON.
How each ticket reaches your repository, on its own branch with a draft pull request, is covered in Clear a ticket backlog.
Good fit
- Teams that want proof for each acceptance criterion, not a statement that the work is done
- Repositories with build, lint, typecheck and test commands to hold changes to
- Reviewers who want to read the diff next to the criterion it satisfies
Not the right fit
- Gates are only as strong as the checks you configure, so a repository without tests has fewer hard gates to rely on
- The reviewer and the verifier work from your acceptance criteria, so vague criteria give them little to confirm
- It installs on a Linux server you operate
- Status today: Onboarding founding customers
Frequently asked questions
In Reqursor Development a ticket ships only when the verdict says ship, every required criterion has file and line evidence and every check is green. An independent verifier goes over the change before that. Otherwise the ticket goes to Needs review with a report of what was tried.
The agent gets a limited number of attempts to fix it, and every attempt goes through the same checks again. If it still cannot make everything pass, nothing is committed and you get an escalation report that explains what was tried and why it stopped.
Yes. The verifier is a fresh, independent agent. It sees only the diff and the ticket, never the implementer’s reasoning, and works out again whether the work is done. Its session is separate from the one that wrote the code.
Yes, your team keeps the final say. Large or sensitive diffs wait for a person to approve them, and every criterion is linked to the diff lines that satisfy it, so the reviewer can check the evidence next to the code.
Related pages
Use case
Clear a ticket backlog
Your developers write tickets with acceptance criteria. Reqursor Development implements them in your repository and, by default, opens a draft pull request for your team to review.
Comparison
Versus editor AI assistants
An editor assistant works inside a developer’s session. Reqursor Development takes tickets from a queue, runs on a Linux server you operate and ships only through your checks and evidence.
Use case
Security sensitive teams
Reqursor Development installs on a Linux server you operate. The agent runs in a sandbox that never sees your data directory, your license or your real API keys, and your code goes only to destinations you configure.
See it on your own codebase
See Reqursor Development work through real tickets on a repository like yours, in a 30-minute call with the people who build it.