devwithjevJev-Powered QA Agent Tests Apps
Stencil rebuilt a QA agent around Jev: a frontier model plans tests, Jev chooses browser actions, and vision handles visual checks.
View on X time28.3 seconds
Dan Leshem@leshemco๐
we rebuilt a QA agent around @typesafeaiโs Jev at Stencil.
a frontier model plans the test. Jev chooses browser actions. vision handles the visual checks.
here it signs in, opens an app, triggers the paywall, and checks a spacing fix.
5 checkpoints passed in 28.3 seconds. every checkpoint has a screenshot, every decision has a trace.
i can see Jev driving decisions across our software factory: which issues to prioritize, what to test, when to retry, and when a change needs human review.
The agent signs in, opens an app, triggers the paywall, and checks a spacing fix. Each checkpoint has a screenshot, and each decision has a trace.
Also filed under Agents & browsers
- Local Jev Decision Model for Claude Code
jev-gate provides a local Jev-compatible decision model as an assistant for Claude Code, with a Bash gate, Stop evidence gate, decide MCP tool, and eval harness.
- Chinese Voice Browser
CatJuly's project is a Chinese voice browser built with Jev.
- Jev-Powered Stealth Browser Agent
CloakBrowser-Agent is a Jev-powered stealth browser agent. TypeSafe Jev decides each step in ~0.3 s, and CloakBrowser carries it out like a human.
- Jev Playwright Test Driver
Oliver Stenbom says Jev struggles to run Playwright tests fully autonomously because of its text/classification approach. He reports that under 50% of their test suite failed with Jev as the driver.