It is 100% compatible with the APIs and all the features of browser-use (https://github.com/browser-use/browser-use) — but rebuilt from the ground up for modern TypeScript environments.
We really like browser-use and started with it, but our tech stack is entirely TypeScript/JavaScript, so we decided to build something that fits more naturally into that environment. We rewrote it with:
- Strong and explicit TypeScript typings
- Full API + feature parity with browser-use
What Browsernode does: It enables LLMs and agents to browse the web using natural language or function calls. It extracts DOM elements (buttons, inputs, etc.), xPaths, and page structures into prompt-friendly JSON.
What it is:
- A browser agent library
- Built for custom LLM workflows, RPA tools, or automation pipelines
- Designed with dev-first principles and great DX
Example tasks (via prompt or code):
- Write a letter in Google Docs and export it as PDF
- Find the cheapest one-way flight from Singapore to Tokyo on Aug 25
- Navigate from “Banana” to “Quantum Mechanics” on Wikipedia via clicks
Try it live or check out the demos: https://browsernode.com Demo video: https://s3.browsernode.com/google_doc_pub.mp4
We’ve been using it internally for over 4 months and felt it’s time to share. Would love feedback from anyone building LLM agents, RPA tools, or custom browser automation.