Stagehand

Stagehand (stagehand.dev) is an open-source browser-automation framework built by the Browserbase team. It sits on top of Playwright and adds three small but powerful AI primitives — act() (do something described in natural language), extract() (pull typed structured data from the page), and observe() (ask the page what elements exist for a goal). Together they make it dramatically easier for an LLM agent to drive real web apps without hand-writing selectors that break the moment the DOM changes.

Unlike pure-vision browser agents, Stagehand is deliberately code-first: TypeScript and Python SDKs, normal Playwright fixtures, and unit-testable functions. Developers describe the intent ("click the second result and extract the price as a number") and the framework handles selector hunting, retries, and structured-data parsing. This is the same pattern Browserbase has been pushing for cloud-hosted browsers — Stagehand runs locally for development and on Browserbase's infra for production.

Pricing is bundled into Browserbase's plans rather than published separately for Stagehand: the framework itself is open source (MIT). You can run it free against your own Playwright browsers, or pay Browserbase for managed cloud browsers, stealth modes, captcha handling, and concurrency. Costs scale with browser-session minutes; check Browserbase's current pricing page before committing for production.

Where Stagehand wins: testing dynamic JavaScript-heavy sites, building agent skills that need to interact with apps without APIs (think internal tools, vendor portals, scraping pipelines), and replacing brittle Playwright suites where DOM changes constantly break tests. The Playwright-compatible base means existing test infrastructure ports over. Where it falls short: it's a developer tool, not a no-code platform — non-engineers will struggle. And complex multi-tab flows still require careful state management; Stagehand doesn't magic away every browser-automation gotcha.

Best for: engineering teams building agent products that need reliable browser actions, QA teams modernizing E2E tests, and developers building scraping or automation systems against modern JS-rendered sites.