Stagehand is for developers who want browser automation they can read and test, with a model filling in the steps that break scripts. It is MIT-licensed, maintained by Browserbase, and since January 2026 available in six languages.
Key Takeaways
- Code first, AI where it helps. — Stagehand lets you script a workflow and hand the hard steps to a model through act, extract and observe calls, with an agent mode for open-ended tasks.
- v3 dropped Playwright. — Since October 2025 Stagehand talks to Chrome over the DevTools Protocol, works with Puppeteer or any CDP driver, caches discovered actions, and is 44% faster than v2 on Browserbase's own tests.
- Every major language. — Since January 13, 2026 there are SDKs for TypeScript, Python, Go, Ruby, Java and Rust on one canonical implementation. The SDK is MIT and free; you pay for the model and the browser.
What does Stagehand do?
Stagehand sits between a scripted test and an autonomous agent. You write the workflow in code; where a selector would be brittle, you ask the model to act on, extract from or observe the page in plain language. An agent mode handles open-ended tasks with any model. Stagehand v3 (October 29, 2025) removed the Playwright dependency in favor of direct Chrome DevTools Protocol control, added caching of discovered elements and actions so repeat runs skip model calls, and a self-healing layer for changed pages.
On January 13, 2026 Browserbase shipped a canonical implementation with SDKs for Go, Ruby, Java and Rust alongside TypeScript and Python. Releases are frequent: the October 2026 updates added page predicate waits, authenticated CDP gateways, nested cross-origin iframe handling and support for the newest Claude models (release notes).
Scorecard: 16/20
Four criteria, 1 to 5 each: how much of the agent loop the product supplies, the hosted browser infrastructure, openness (license, self-hosting, model choice), and published cost. These are editorial assessments of documentation and pricing pages, not benchmark runs. They match the Browser Automation for AI Agents guide.
| Criterion | Score | Why |
|---|---|---|
| Agent | 4/5 | AI primitives for single steps plus a model-agnostic agent mode; self-healing when pages change. |
| Infrastructure | 2/5 | A library, not a host: runs a local browser or a Browserbase session. |
| Openness | 5/5 | MIT, six languages, any CDP driver and any model. |
| Cost | 5/5 | Free SDK; cached actions cut repeat model calls. |
What does Stagehand cost to run?
The SDK is free under MIT. Running it costs two things: model tokens, from your own provider or through Browserbase's model gateway, and a browser. A local Chrome costs nothing; a hosted Browserbase session starts with the $20 a month Developer plan (100 browser hours, then $0.12 an hour) as published in October 2026. Because v3 caches actions, a workflow that repeats calls the model less often than an agent that re-plans every run.
Who it fits, and who should look elsewhere
- Fits: engineering teams that want automation in code review and CI, with AI only where pages are unpredictable, in TypeScript, Python or another of the six supported languages.
- Look elsewhere: if no one will write the workflow and you want a goal-driven agent (Browser Use), or if you only need an LLM client to drive a browser through MCP (Playwright MCP).
Stagehand alternatives
- Browser Use: an autonomous agent library with its own cloud.
- Browserbase: the hosted browser Stagehand is designed for.
- Playwright MCP and others: see the full guide.
Related questions
- Which browser infrastructure should developers use for AI agents?
- Browser Use, Browserbase or Stagehand: which fits?
- When should you use an AI agent instead of scripted automation? (partner site CTAIO)
- Which agent orchestration framework should run the agent that drives the browser?
Sources
- Browserbase, Launching Stagehand v3 (Oct 29, 2025).
- Browserbase, Stagehand is now available in every language (Jan 13, 2026).
- Releasebot, Stagehand release notes (through Oct 30, 2026).
- eCorpIT, Cloud browsers for AI agents 2026 (Oct 22, 2026).
- Toolradar, Browserbase pricing (prices confirmed Oct 30, 2026).
Is Stagehand free?
Yes. The Stagehand SDK is open source under the MIT license. You pay for the language model it calls and for the browser it runs in, which can be your own Chrome or a paid Browserbase session.
Does Stagehand require Browserbase?
No. Stagehand can connect to a local browser or any CDP endpoint. Browserbase adds hosted sessions, saved logins, CAPTCHA solving and replay, and a model gateway Stagehand can route through.
What changed in Stagehand v3?
Released October 29, 2025, v3 talks to Chrome directly over the DevTools Protocol instead of through Playwright, supports Puppeteer and other CDP drivers, caches discovered elements and actions, and adds a self-healing execution layer. Browserbase reported it as 44% faster than v2 on its benchmark suite.
Stagehand or Browser Use?
Stagehand when you know the workflow and want most steps in code, with AI for the fragile parts. Browser Use when you want the agent to plan every step from a goal. Our comparison with Browserbase covers the trade-off in detail.
Ready to Find the Right AI Tools?
Browse our data-driven rankings to find the best AI tools for your team.