Agentic browser QA — Claude in Chrome
Run the v0.3 QA beats by pointing the Claude-in-Chrome extension at the live dashboard — an agentic complement to the Playwright suite.
What it is#
Instead of scripted DOM asserts, you paste a natural-language prompt into Claude in Chrome (active on app.controlum.pw) and it performs the beat — discovery, live-stream-no-freeze, R2.3 prompt→PTY, kill-API-no-gap, version — then reports back, with a screenshot. You verify its report against ground truth.
Playwright vs Claude in Chrome#
- Playwright — scripted, deterministic, runs in CI, gates merges, counts toward the ≥98% coverage target.
- Claude in Chrome — agentic, on-demand, best for UX/render realism and the pre-demo "does it feel right to an agent operating it" pass. Additive; does not count toward the coverage number.