Introducing Stagehand v4: the SDK for browser agents. Playwright was built for testing, we built Stagehand for your agent: with improved context management, self-healing actions, and iframe support.

Stagehand now lives in the browser as an extension, bringing the local experience to remote browsers. Being in the browser means lower round trip time, which makes actions execute faster and minimizes potential CDP race conditions.

We benchmarked v4 extensively and found that it’s performs 2x faster than Playwright, while being ~80% more token efficient.

We partnered with @vercel, @mastra, @crewAIInc, and @LangChain on browser agent templates and guides for building with Stagehand v4. Read the launch blog here:

Build and quickly deploy browser agents with Vercel’s Eve framework. Eve brings a filesystem, sandbox, and now a full remote browser to interact with the web.

Build browser agents in TypeScript with Mastra. Get detailed traces and maximum observability through the Mastra & Browserbase dashboards.

Build browser agents with LangChain DeepAgents. Build long-horizon agents that can use the internet and improve them with extensive logs, traces, and browser session replays.

Build multi-agent crews of browser agents with CrewAI in python. Bring reliable browser interaction to your enterprise ready agents.

Give your agents access to the web with Stagehand:

This looks great! And thank you for not sitting statically in front of beige furniture

only fun videos here!

“this shit cooks” - my agent

valid

I’ve always dreamed of working for Stagehand IV I joined when it was only Stagehand Junior

3 generations of stagehand later

we just killed playwright

thank you @endpoint for giving us stagehand

@endpoint one of the goats

@endpoint in sameel we trust

Evals!

evals are a must

now get to work on v5 @mcguiresean_

@mcguiresean_ in v5 we'll be faster than cdp

LFGGgg!! Much needed update. Time to unleash it.

it's been a long time coming ...

Finally it’s here…

my goats

the original

incredible! so excited to see

excited for Mastra's agents to use stagehand in production

sick. also the videos are really getting good!

video quality directly correlated with product quality

love this video + congrats on the launch! @kylejeong +team

huge

huge and faster than ever

You absolutely love to see it! 🤝

LFG!!

it's finally live

Beautiful work! I’ll be (stress) testing your implementation with our PDF SDK - we heavily rely on workers, and iframe communication so a perfect “complex” candidate to test your tech

Stagehand + Browserbase > Everything else

i’m so excited!!!!

we need to enable stagehand tools in @composio

big day!

appreciate your support!

Glad you joined @rtrvrai 's philosophy finally. Well I bet you can't match our cost efficiency.

The iframe support is the part I would have led with. That is where most browser agent runs quietly fail, the action lands on the top document while the target sits two frames down. Curious how self healing handles structure that changes mid run rather than between runs.

@yuris Dude has the forth in his name

Congrats! This reminds me I need to tell my agents to stop using Playwright and use Stagehand.

they'll be much better off with Stagehand! we've got an MCP tool that we've found useful in our integrations:

Thanks!

as of recent my agents love stagehand

stagehand loves your agents too

Playwright tests the happy path. Stagehand survives the browser fighting back.

stagehand was built to be resilient

GG @pk_iv!!

iv!

the fourth and best version of Stagehand

And arguably pk

congrats on the launch! looks awesome!

Congrats on the launch @kylejeong !

Playwright was great for testing, but an SDK built specifically for agents makes a lot of sense

🐐

Self-healing actions are useful, but the failure report matters just as much When the page changes, operators need to know what the agent tried, what it changed, and where human review is still required