How Vela made engineering reviews 3× faster with cloud agents
VelaVela is an AI scheduling assistant for professional services and executive search firms. Teams CC Vela on an email thread like a human EA, and it handles the back-and-forth across email, SMS, WhatsApp, Slack, Teams, and phone. It follows up with non-responders, handles time zones and rescheduling, and coordinates multi-party interview panels. Vela schedules thousands of meetings a week.
Vela’s engineering team uses Replicas to review and test pull requests, give reviewers a visual summary of changes, and triage production alerts. Engineers and other teammates also start investigations and ship fixes from Slack.
Visit Vela“With Replicas, everyone on the team investigates issues and pushes new features, all with reduced review times that allow us to provide our customers a better product, faster. At the same time, we keep costs low by using our own harnesses and models. In hindsight, I can't see how we did things before Replicas.”
3×
faster engineering reviews
Two independent reviews before a human opens the diff
Vela puts pull requests through a set of checks before an engineer picks them up. Across its frontend and backend repositories, a Claude agent reviews each change against the team’s standards while a Codex agent takes an adversarial pass and tries to break it.
A QA agent generates and runs end-to-end tests for each pull request. When reviewers want a faster way into a change, a visual recap agent produces a before-and-after summary of the diff on demand. The review starts with independent findings and test results rather than a bare patch.
From production alert to an owned fix
Vela applies the same approach to incidents. When Sentry or Grafana raises an alert, a triage agent investigates whether it needs attention. It files a GitHub issue and assigns an owner when it does, and can open a small pull request itself once the root cause is confirmed.
The team also uses Replicas from Slack to investigate production issues, trace bugs, add migrations, and open pull requests. The work can start in the conversation where someone noticed the problem.
3× faster reviews
Every pull request gets reviewed by two different models and tested end to end; reviewers can request a visual recap before reading the diff. Vela’s engineering reviews are now 3× faster, and the team deploys changes more quickly. Alerts that need fixing can become owned issues or small pull requests rather than stopping at a notification.
