Stagehand

Stagehand is an open-source SDK built by Browserbase for creating AI agents that browse and interact with websites. Rather than writing brittle selector-based scripts, developers describe what they want done in natural language, and Stagehand translates that into browser actions through three core APIs: Act, which executes commands like clicking or filling forms; Extract, which pulls structured data out of a page using a schema; and Observe, which lets the agent understand what is currently on screen before deciding what to do next. Stagehand is positioned as an alternative to Playwright for agentic use cases specifically, rather than traditional test automation, and the maintainers report it running about twice as fast as Playwright in production while using roughly 80% fewer tokens for page understanding (7.4k versus 35.7k tokens in their benchmark). It also supports WebMCP, clipboard access, batched commands, iframe and Shadow DOM handling, and OpenTelemetry tracing for observability, and can run as a browser extension to cut latency. The SDK is available for TypeScript, Python, and Go and is installed via npm, with documentation at docs.stagehand.dev and a Discord community. It integrates naturally with Browserbase's hosted browser infrastructure for production deployments, though it can also run against locally controlled browsers. As an actively evolving open-source project, some advanced behaviors may change between releases.