Lark

About Lark

Lark, built by ex-Stripe engineers, runs UI, API, CLI, and mobile tests on a continuous cadence rather than only on commit — the framing is 'every surface, every minute', with the team watching green while Lark watches red. That cadence is the product difference: most suites only tell you something broke when someone pushes, whereas Lark keeps checking against production and staging so a third-party outage, an expired credential, or a config drift surfaces on its own. The dashboard shown on the site reports over 12,400 runs completed in a day at a 99.6% pass rate with 31 tests auto-repaired, which points at the second core feature — self-healing. Tests adapt when you revamp a dashboard or introduce a new API instead of breaking, and Lark alerts you to the change rather than silently failing. It is built for developers rather than a separate QA org, with first-class support for coding agents like Claude Code, Cursor, and Codex, so tests can be authored and maintained from inside the agent workflow that generated the code. When something does fail, Lark produces reproducible debugging artifacts: downloadable test scripts, logs, screenshots, and videos. Coverage spans UIs, APIs, SDKs, and async workflows.

Key Features

Lark Pros & Cons

Tags