Your agents ship faster than you can review.
You can't trust what you haven't read, so you read everything. Aonyx checks your agents' work and hands you only the things a human has to decide.
One queue that tells you when you're needed.
Everything your agents did lands in one queue, already checked. What passed cleared itself — nothing to read. What needs a human call waits at the top, so you look once, not all day.
Workflows you define, runs that finish without you.
Each workflow is a piece of your process, defined by you: the steps, the checks, and what happens when one fails. This run hit a failed check and re-ran — not platform magic, just your own rules being followed. That's why you can trust what comes out.
The information you need, at the moment you decide.
Work moves through your org checkpoint to checkpoint — a plan, a PR, a release — carried by workflows in between. Each checkpoint shows just the information that decision needs; you choose where the work goes next: onward, back for another pass, or something else entirely.
Add gust smoothing to the autopilot
Prove each version of your workflow beats the last.
Every run adds to your workflow's track record. Change the workflow, and the runs that follow show whether it really got better — cheaper, fewer retries, trusted more — so "we improved it" is a number, not a feeling.
Agents you can trust, measure, and improve.
Help build it. Get it first.
Honest status: there's nothing to install yet. We're aiming to put the first release in this list's hands in late August. Until then, we'll ask you the questions we're wrestling with — like whether writing workflows in TypeScript works for you — and your answers will shape what ships.
Built by Jan David (ex‑Rust Foundation) and Martijn (ex‑Palantir).
Expect the occasional real question from us — and first access when it ships.
