Phase 03 · the engine that runs itself
Your Sims test every release. Unchanged screens replay with zero AI — fast and free enough to gate every build. When a button moves, they re-find it by meaning and keep going. When something truly breaks, they file a grounded bug — never a silent false-green.
The AutoSim critical path
Describe a flow in plain English, and Klavity turns it into a real, reusable test. From then on it replays with no AI on every build, adapts when your UI shifts, and files a grounded bug only when the goal genuinely can't complete.
Say the flow (or record it). A Sim drives the browser and Klavity saves it as a test you own.
Unchanged UI replays with zero AI: fast, cheap, and repeatable enough to gate every build.
When your button moves, the Sim re-finds it by what it means and keeps going — flagging the change, not failing.
If the goal genuinely can't complete, it files a grounded finding instead of quietly passing green.
One dashboard: what passed, what healed (with the before→after), what broke — each with a scrubbable replay.
Why AutoSim is different
Most test automation pays for maintenance with false confidence. AutoSim flips it: heal the noise, surface the signal, and stay cheap enough to run on every commit.
When something moves, Klavity re-finds it by what it means — first from memory, then by its on-screen label, then by looking at the page. It's diagnosis-first: it adapts only when the goal still completes, so a real break is filed, never papered over.
Unchanged screens replay from memory with no AI at all — repeatable and cheap enough to run on every release. AI is spent only to set up, adapt, or diagnose — never on a passing run.
Your Sims — personas built from real customer calls — are the oracle. A run isn't "did the script pass?" but "could this person actually finish what they came to do?"
Every flow becomes a real, standard test you own — keep it, export it, run it anywhere. AutoSim is open-core: the engine earns its keep instead of trapping you.
One thread, all the way
A human snaps a bug, AI Sims catch them for you, then those same Sims run themselves — testing every release and healing when your UI changes.
Questions
When something on the page moves, AutoSim re-finds it by what it means — first from memory, then by its on-screen label, then by looking at the page — and keeps going, flagging the change (e.g. the checkout button moved) instead of failing.
No. Healing is intent-verified and diagnosis-first. If the goal genuinely can't complete, AutoSim files a grounded finding rather than quietly passing green — so it never produces a silent false-green.
Unchanged screens replay with no AI and the same result every time, making it cheap and fast enough to run on every release. AI is only spent to set up, adapt, or diagnose.
Yes. AutoSim is open-core and turns each flow into a real, standard test you own — keep it, export it, run it anywhere, with no lock-in.
Phase 03
Continuous testing that monitors your product like a QA team that never sleeps — and never a silent false-green.