Built for QA people who'd rather describe tests than maintain them.
Orbis QA started as one engineer's career-pivot project — a way to ship a real AI/ML product end-to-end rather than read another tutorial. It's grown into a tool we think every QA team should have access to.
Every QA tool we tried was built for engineers — Playwright, Cypress, Selenium. Pretty APIs, but they all assumed you'd write code, maintain selectors, fight flakes, and re-author tests after every UI redesign. The actual SQAs in our orbit spent more time keeping the test suite alive than testing the product.
We wanted a tool where you could describe a flow in English and an agent would just do it — recover from glitches, take screenshots as it goes, and tell you exactly what broke. The LLM stack finally got good enough for that in 2026. Orbis QA is what we built.
Orbis QA is free during launch — every feature is unlocked. Paid Pro and Team plans come later. They unlock optional add-ons (scheduled runs, cloud sync, shared libraries) without taking anything away from Free.
Four opinions that shape every decision
Local-first by default. Scenarios, runs, screenshots — none of it touches a server we operate. The Pro tier will offer optional cloud sync; the Free tier will always run offline.
We never sit between you and your LLM provider. You pay the model directly, the key is encrypted in your OS keychain, Orbis never proxies a single token.
QA is messy. Recovery sub-loops, flake detection, multi-role sessions — we're shipping the things that turn 'it works on my laptop' into 'it's safe to run unattended overnight'.
No selectors. No fixtures. No CI rewrites every time the UI changes. Orbis is built so a scenario you wrote in 2026 still runs in 2028 against a completely redesigned site.
Talk to us
We answer everything personally — bug reports, feature requests, "this didn't work for me", "we'd pay for X". The contact form is the fastest route; email works for anything private.