AI-assisted bug fixing that opens the pull request
Pick an error and the Conductor drafts the fix as a pull request your team reviews and merges. It works on a scoped, double-scrubbed context, runs on your own model key, and never auto-merges.
AI that opens a fix PR is table stakes now. What matters is how, where and on whose terms it runs. Condux draws the boundary tightly: draft PRs only, credential isolation, a full audit trail, and your own model.
The safe path from error to PR
- Scoped context
The stack trace plus code mappings, double-scrubbed for PII before it ever reaches the model.
- Draft PR only
The fix lands on a new branch as a draft PR authored by the bot. Never an auto-merge.
- Credential isolation
A least-privilege GitHub App with short-lived tokens. The model never sees the token.
- Full audit trail
Every run records its model, token counts and outcome, verifiable in the source.
It proves the fix worked
After the PR merges and deploys, Condux watches that exact issue against its real per-hour stats and auto-resolves it with evidence once it stays silent for a set window. It fixed it, and it can show that it stayed fixed.
Frequently asked questions
Does the AI merge code on its own?
No. The Conductor only ever opens a draft pull request. A human on your team reviews and merges it through your own CI.
Can the model see my repository token?
No. Repository access is a least-privilege GitHub App with short-lived installation tokens, and the token is never sent to the model.