BrowserBook is an AI-powered IDE for building fast, reliable browser automations. It pairs a Jupyter-style notebook with a context-aware AI coding assistant and an inline browser, enabling developers to write, test, and debug automations all in one place. With built-in auth management, data extraction, screenshot tools, and API deployment, you can go from idea to production-ready automation in minutes, not hours.
Hey Product Hunt! 👋 I’m Chris, co-founder of BrowserBook here with teammates Jorrie and Evan.
We’re excited to share BrowserBook, an AI-powered IDE for building reliable, deterministic browser automations. It combines a notebook-style interface, an inline browser, and a context-aware coding assistant to help developers write, run, and iterate on Playwright scripts faster with built-in tools for screenshots, data extraction, and managed auth.
Why we built this Jorrie and I started a healthcare automation company last summer, where we spent lots of time building back-office workflows (think insurance followup, medical coding, etc.) with custom browser agents. During this time, we ran into four major issues with browser agents:
💰 They’re expensive - we burned through tokens with all the workflow context we needed to provide 🐢 They’re slow - all that context let to super high latency on each request 😬 They’re unreliable - even with detailed instructions and tools, agents still drifted on multi-step tasks 🐞 They’re impossible to debug - when they drifted, there was no clear root cause; we often resorted to trial-and-error to resolve issues.
We found ourselves giving our agents scripts to execute at each step of the workflow, and eventually came to the conclusion that scripting was just the better solution. We decided to put the agent at the coding step rather than the execution step, using LLMs to rapidly write scripts based on DOM context, and ultimately built BrowserBook as the tool we wished we had.
So what's included? The core IDE experience is purpose-built for web automation; some of its features include:
🌐 A fully-interactive inline browser instance that you can run Playwright automations against with the click of a button 📒 A Jupyter-notebook-style interface, so you can just run parts of your automation instead of the whole thing every time 🤖 A context-aware coding assistant so you don’t have to go digging for selectors 🤝 Helper functions for common actions like taking screenshots, data extraction, and managed auth profiles supporting username/password and TOTP for authenticated workflows
When you’re done building, you can run your automation manually through the IDE or via API for use in external agents or apps!
Our initial use case was healthcare automation, but if you’re building automated tests, writing browser-based RPA workflows, web scraping, or have an agent that needs to perform deterministic actions on the web, give us a try - we’d love to hear what you think!
Finally a tool for this that is not just a low-code drag and drop wrapper. Love the dev-first approach. How do you guys handle stuff like Cloudflare turnstiles or captchas in the inline browser?
Since the output is Playwright scripts, are the generated scripts fully standard and compatible?
Congrats on shipping BrowserBook!
The “agent at the coding step, not the execution step” idea really resonates — I’ve hit the same issues with browser agents being slow, expensive and impossible to debug.
The inline browser + notebook flow looks perfect for iterating on automations.
Excited to try this for some internal RPA-style workflows 🙌
I’ve been tinkering with BrowserBook since the alpha launch, and it’s been awesome. Writing browser automations used to be the worst. I either had to tediously write them by hand or use a non-deterministic browser agent. Browser book gives me the best of both, all the power of AI with all the reliability of hand written code. There are many more things I want to build with it. Super excited to see how everyone uses it. Congrats team!
I've been waiting for an IDE that has an actual browser in it with the DOM. Despite all the LLM tools these days, I still spend so much time switching between the console, the browser, and vscode. Really cool to finally see it here!
Congratulations on the launch, team! Excited to give this a whirl, it looks awesome.
Do you guys plan on doing anything at all with browser agents? I feel like they’re useful for adapting on the fly for intermittent things like popups or modals that would get in the way of a normal workflow
Such a great product with a unique approach. We've been working with BrowserBook, and they know the ins and outs of this space. If you need a way to do repeated browser automations, these guys have cracked it.
Congratulations on the launch and I like the idea of being able to set up complex automations across any website. I enjoy using GumLoop but I sometimes feel like it's not as customizable as I would like it to be.
Nice work boys. What's the deployment process like? Is there an API you can use to call automations you build (apologize if this is on the launch video). Use case I have: running a deep search on an address (usually real estate platforms). Currently do manually. Tried to automate with browser-use but it was pretty flaky and didn't have time to figure out the API
Seems super helpful for building end-to-end tests and evals for webapp-based LLM products without having to rearchitect/reinstrument your whole app. Gonna give it a try and see if I can get a playwright automation going
I'm constantly getting fun browser based ideas w/out any desire to rip off that implementation band aid. May even use this to solve some of my personal pains
Hi Product Hunt, other co-founder of BrowserBook here! Chris has already explained some of the reasons why we built BrowserBook, so I'm going to share what makes us different from other web automation platforms and why you should give us a try.
The automation software we've seen falls roughly into two categories:
Software SDKs where you need to build it all yourself.
Full automation platforms that tout low or no code automations.
BrowserBook is the sweet spot between these two options.
Don’t know where to start with authentication, hosted browsers, data extraction, session recording, etc.? BrowserBook sets all this up for you with a simple interface that abstracts away the complicated parts so you can focus on writing scripts.
Nervous about using no code platforms to write your production workflows because it’s a startup and when they go out of business in three months all your carefully crafted prompts will be wasted effort? Use us instead! When we go out of business in three months, what you’ll be left with is a bunch of working playwright code.
Hey Product Hunt! 👋 I’m Chris, co-founder of BrowserBook here with teammates Jorrie and Evan.
We’re excited to share BrowserBook, an AI-powered IDE for building reliable, deterministic browser automations. It combines a notebook-style interface, an inline browser, and a context-aware coding assistant to help developers write, run, and iterate on Playwright scripts faster with built-in tools for screenshots, data extraction, and managed auth.
Why we built this
Jorrie and I started a healthcare automation company last summer, where we spent lots of time building back-office workflows (think insurance followup, medical coding, etc.) with custom browser agents. During this time, we ran into four major issues with browser agents:
💰 They’re expensive - we burned through tokens with all the workflow context we needed to provide
🐢 They’re slow - all that context let to super high latency on each request
😬 They’re unreliable - even with detailed instructions and tools, agents still drifted on multi-step tasks
🐞 They’re impossible to debug - when they drifted, there was no clear root cause; we often resorted to trial-and-error to resolve issues.
We found ourselves giving our agents scripts to execute at each step of the workflow, and eventually came to the conclusion that scripting was just the better solution. We decided to put the agent at the coding step rather than the execution step, using LLMs to rapidly write scripts based on DOM context, and ultimately built BrowserBook as the tool we wished we had.
So what's included?
The core IDE experience is purpose-built for web automation; some of its features include:
🌐 A fully-interactive inline browser instance that you can run Playwright automations against with the click of a button
📒 A Jupyter-notebook-style interface, so you can just run parts of your automation instead of the whole thing every time
🤖 A context-aware coding assistant so you don’t have to go digging for selectors
🤝 Helper functions for common actions like taking screenshots, data extraction, and managed auth profiles supporting username/password and TOTP for authenticated workflows
When you’re done building, you can run your automation manually through the IDE or via API for use in external agents or apps!
Our initial use case was healthcare automation, but if you’re building automated tests, writing browser-based RPA workflows, web scraping, or have an agent that needs to perform deterministic actions on the web, give us a try - we’d love to hear what you think!
How do I use it?
You can download for Mac at our website: https://browserbook.com
Cheers,
Chris