A proactive assistant for engineering teams
When an issue, failed build, or review thread matches rules you have approved, Afe starts the work. It returns with a checked result that is ready for review—or one clear question when it needs your decision.
Afe traced intermittent auth failures to a timeout mismatch between the gateway and identity service.
Runner image change linked to 3 failing jobs · 11 min
4 owners, 2 code changes, and a runbook gap found
Afe works with the tools your team already uses
How Afe works
You choose which kinds of work Afe may take on. When matching work appears, Afe starts, tracks each remaining step, and asks you only when it needs your judgment or approval.
A release blocker, incident follow-up, review thread, or failed check can signal work Afe may handle. Afe starts only when the job matches rules you approved.
Your tools · approved rules · no extra promptAfe pulls together the ticket, code, logs, recent changes, and related discussions, then prepares the change in isolation and runs the checks your repositories require.
Read · change · testIf CI takes twenty minutes or a reviewer replies tomorrow, Afe resumes the same job from its last saved step. You do not have to restart it.
CI · reviews · follow-upsWork continues after the code
A patch is rarely the end of the job. Afe waits for CI, follows up on straightforward review comments, and keeps the work linked across repositories and tickets. When a job is waiting, Afe saves its progress and resumes when something changes.
See what Afe is allowed to doGateway and identity fix · allowed by your rule
Progress saved · Afe will resume automatically
Afe found conflicting service conventions and laid out the trade-offs
Both diffs, test results, rollout notes, and one open trade-off are attached
64 contract and integration tests passed against both commits
CI logs, gateway traces, and the last identity rollout pointed to the timeout mismatch
A clear progress record
Each job has a plain record of the sources Afe read, the changes it made, the checks it ran, and any limits it hit. If something fails, you see that too.
Built for trust
A ticket appearing does not mean Afe can act on it. You decide which kinds of work, repositories, and actions Afe may handle.
Afe keeps the ticket, commit, approval, and test results together, so you can review a change from one record.
Production changes, destructive actions, and security-sensitive work stop for your approval—even when Afe thinks it knows the answer.
One job across connected tools
Afe can move a job from ticket to code to CI to review without losing its history. Afe may use different tools for each step, but you still see one record and one place to decide.
Early access
It might be a release blocker across two services, incident follow-ups, a CI regression, or review threads from last week. Tell us which work you would want Afe to handle first.