Explore — alternate browsers

One-time spike before 1.0: could a faster browser backend beat Playwright on mass recipe e2e (leased URL, many workers)?

Decision for 1.0: stay on Playwright. Revisit when real perf dogfood needs throughput beyond Node CDP bindings — not because Playwright “already has CDP.”

Project	Perf angle	Why not in 1.0
Astral (Deno)	Lean Deno browser automation	No Node/Vitest dogfood path without a separate runner
Thirtyfour (Rust / WebDriver)	Native Rust client	WebDriver stack + Node bridge; unclear win vs CDP for our recipe shape
Chromiumoxide (Rust / CDP)	Low-level Rust CDP for raw throughput	Interesting because it is CDP — potentially faster than Playwright’s Node layer. Deferred for Node↔Rust integration cost in the Vitest/Playwright recipe path until there is concrete perf dogfood

A thin recipe that only returns { url } could sit under any browser client later. That is not worth a first-class adapter until something measures better than Playwright on real suites.

Do not block 1.0 on shipping these.

Related: Vitest Browser Mode (component-in-browser track — also not recipe-e2e).