Stagehand
Browser-automation SDK for AI agents with Playwright-style APIs and self-healing actions
💬 What is Stagehand?
Stagehand is an open-source SDK for building browser-based AI agents. It is designed for developers and engineering teams that need production-grade web automation that combines natural-language instructions with deterministic Playwright-style actions. The framework runs as a browser extension to minimize round-trip latency and provides self-healing selectors, WebMCP support, batched commands, deep locators for iframes and Shadow DOMs, and built-in OpenTelemetry tracing. It exposes SDKs for TypeScript, Python, and Go, and is maintained by Browserbase. Compared with general-purpose testing tools like Playwright, Stagehand is agent-first, optimized for token efficiency and resilience to UI changes.
Who it's for
Best for: Developer and platform-engineering teams building production browser-automation agents
- Building LLM-driven agents that interact with live websites
- Resilient web scraping and data extraction that survives UI changes
- Automating multi-step browser workflows with natural-language instructions
- Adding browser-control tools to AI agents and assistants
- Production web automation with observability and access controls
Pros & cons
- Established open-source project with 23.7k+ GitHub stars and active community
- Multi-language SDKs (TypeScript, Python, Go) with consistent API
- Production-oriented features: OTel tracing, domain allowlist/blocklist, WebMCP
- Self-healing actions reduce breakage when websites change
- Runs as a browser extension for lower round-trip latency than remote drivers
- Currently at v4.0.0; recent major-version changes mean APIs may continue to evolve
- No public pricing page on stagehand.dev for managed or hosted usage details
- Requires programming skill; not a no-code tool
- Comparison benchmarks are self-reported on the project's own site
Key features
- Playwright-style API (goto, click, locator, screenshot)
- Natural-language act, observe, and extract primitives
- Self-healing selectors that adapt when page structure changes
- WebMCP support for agent tool integration
- Clipboard access and batched command execution
- Deep locators for nested iframes and closed Shadow DOMs
- Built-in OpenTelemetry tracing
- Domain allowlist and blocklist controls
💬 Community sentiment (from public reviews)
No third-party reviews found yet.
💬 Stagehand user reviews (0)
Write a reviewNo reviews yet. Be the first to review!
Badge for Stagehand
OptionalListing is free and never requires a link. If stagehand.dev links back to this page, the listing earns Partner perks: a higher category rank, a Partner badge on its card, and an ad-free direct “Visit” link. Use any link attributes you like — rel="nofollow" is fine.
Frequently asked
Who maintains Stagehand?
Stagehand is built and maintained by Browserbase, with the source code on GitHub (23.7k+ stars at time of writing).
Which languages are supported?
TypeScript, Python, and Go are officially supported, with example code and quickstarts for each.
How does Stagehand differ from Playwright?
Playwright is designed for testing; Stagehand is designed for agents, adding natural-language act/extract, self-healing actions, WebMCP, domain controls, OTel tracing, and token-optimized page context, running inside a browser extension for lower latency.
Does Stagehand require a remote browser service?
No. Stagehand can run against a local browser (e.g., via localBrowser) as well as inside Browserbase's managed browser infrastructure.