Rules your agents actually follow
unerr applies your team’s rules at the moment of the edit, inside the agent’s own hook layer. Not a linter that complains afterwards. A decision made before the change lands.
Free, local, today
Blast-radius guard. Names every call site an edit puts at risk, inside the edit itself.
Convention detection. Finds the pattern that holds across your codebase and applies it where the agent is working.
Retry-loop breaker. Stops an agent repeating the same failing attempt.
Deny and rewrite. Your rules, enforced in the agent's own hook layer, before the change lands.
curl -fsSL https://unerr.dev/install | shNo account. Runs on your machine. Uninstall with one command.
Where paid starts
Everything that runs on your machine is free and always will be — no account, unlimited repos. Anything that needs our servers is the paid plan.
That line is not arbitrary. A single engineer does not want shared policy, audit trails or a fleet view. An organisation will not deploy without them. So the paid side is organisational control, never more capability. You are not buying a better version of the tool.
What isn’t shipped
Org-wide enforcement, meaning blocking and approving across every repository from one place, is not shipped. It is being built. We would rather you knew that now than found out in a pilot.
We don’t rank developers
We measure at team and system level, never per person. That is contractual, not a setting you could turn on.
It is also the only way the numbers stay useful. The moment a metric is used to rate someone, people optimise the metric and the data turns to noise.
Deploying this across teams?
Tell us how your teams are set up and we’ll tell you what applies today and what doesn’t.