An autonomous engineer for pull requests

Finds, fixes, and verifies problems in pull requests — with evidence, not assertions.

Evidence over assertionSafe by defaultOne execution boundaryProvider-neutralIssues become pull requestsOne deployable app

Every fix carries the commands that verified it. Feedback is never treated as truth merely because it came from a human or an AI reviewer. :::

The observe mode inspects and reports and never writes to a target repository. Automatic fixes require confidence, an allowed change-risk class, repository permission, and verification. :::

A single runner package is the only code allowed to run commands or mutate a checkout, locally or inside an isolated sandbox. :::

GitHub, GitLab, Bitbucket, and Gitea adapters on one side; OpenAI, Anthropic, Google, and OpenAI-compatible model providers on the other. The runtime depends on neither. :::

A labeled, repository-scoped issue can be investigated, implemented on an isolated branch, verified, and published as a pull request carrying its acceptance criteria and evidence. :::

A single Nuxt dashboard serves the UI, a typed oRPC control plane, an OpenAPI surface, and authentication from one origin. :::