Introducing BrowserBook Agent
Hey PH, today we're launching the BrowserBook Agent!
We’ve been working on a way to move past manual, step-by-step automation building. The new Agent is a fully autonomous builder designed to take a single high-level prompt and turn it into a complete, deterministic workflow - handling all the technical heavy lifting behind the scenes. See it in action here:
Here is a quick look at what it brings to the table:
One-Prompt Architecture: You provide the goal, and the Agent autonomously builds the entire automation in seconds. It can not only write code, but execute it on your behalf, building a deterministic automation as it works through your workflow.
Dynamic DOM retrieval: The Agent can determine when it needs an updated look at the DOM, and can do so dynamically - so no need for you to go digging for selectors.
Real-Time Problem Solving: Instead of just writing code and hoping for the best, the Agent proactively diagnoses and fixes errors on-the-fly while it’s building.
Deterministic Reliability: At the end of the run, the Agent has completed your workflow and built a deterministic Playwright automation that you can run cheaply and quickly.
Of course, you can still use the agent to make spot edits if you prefer the step-by-step approach. The idea here is to help you go from a concept to a fully functional and cost-effective workflow in a fraction of the time it used to take.
We’d love to hear what you think or answer any questions you have about how the autonomous builder handles complex sites! You can download the latest version now at browserbook.com.
Replies
This is impressive. Going from a single intent → deterministic Playwright workflow is exactly the missing layer between “AI demos” and production automation.
The dynamic DOM retrieval + self-healing during build is especially interesting — that’s where most browser automations usually fall apart. Curious how it handles highly dynamic apps (SPAs, auth-heavy flows).
BrowserBook
@allinonetools_net Thanks Bhavin. Agreed, building flashy demos is one thing, but we're really focused on solving the production use case well, and we think the deterministic approach is the right one in most cases.
To handle more dynamic apps like SPAs, the agent can refresh its DOM context as necessary, so even as the page changes it can get updated information about what's available to interact with. And since it's writing Playwright along the way the end result is an automation that can handle the variability of the DOM.
Re: auth-heavy flows, we have authentication built-in with managed auth profiles, so as long as you have credentials for the applications you need to automate, the agent can retrieve and use them on an as-needed basis (with the user's permission, of course)
@cschlaepfer That makes sense — deterministic runs are huge for trust, especially in prod automation.
The snapshot + auto-repair approach sounds like a solid middle ground between brittle scripts and full autonomy.
Curious to see how this performs on long-running flows with auth + state 👍