Browser Use

Introducing: Browser Use v4 ⚡️ v4 is the best agent we have ever shipped — and now it is one request away. → Real browsers with logged-in profiles, proxies, and a live view → A full sandboxed workspace per agent — upload files, get files back, reuse environments across runs → Conversations that continue — follow up on any run and the agent picks up with full context → No scripts, no selectors — plain English in, results out This is everything we have learned about browser agents, distilled into a single agent. Get your API key at https://lnkd.in/g3eV27-e

We just crossed 100k stars on GitHub, so we’re celebrating by giving away up to $100k 💰 We gave an AI agent a company card. Here’s the challenge: Open Browser Use Cloud. Talk to the agent. Convince it to buy something for you. The agent can browse, reason, and take action on the web. You can ask for anything: • a coffee • a coconut • a car Claim your gift! Only available today from 9am to 9pm PST. Try the challenge here: https://lnkd.in/gFGh8227

We're open-sourcing the browser tools used by Codex, Claude Code, and Cursor. Browser Use CLI 3.0 turns any coding agent into a state-of-the-art browser agent: → Direct CDP control via browser-harness. The agent drives Chrome natively, no brittle abstraction layer in between. → Works with your real Chrome or our cloud browsers. → 6× smaller than Browser Use CLI 2.0, with far fewer tokens per step. Faster runs, lower cost. Everything we've learned from millions of browser-agent runs, now open for anyone to build on. Repo + docs → https://lnkd.in/gYmc-GpM What's the first thing you'd let your coding agent do in a browser? 👇

We launched Fetch Use, the easiest way to scrape any website with the stealthiest browser on the planet. Proxies, cookies, and sessions are all handled for you, so you don't worry about getting blocked. See our blog post for more: https://lnkd.in/g6dQpxiG

Honestly, there is this feeling which can't imagine that this will be higher next week. But I remember the exact same feeling on 12-29 when I saw a similar spike.

We rebuilt Browser Use from scratch (in Rust) 🤯 The best browser abstraction is a tiny browser abstraction 🏴☠️ Old Browser Use: > predefined action spaces > browser abstractions > brittle workflows The model never knew why it failed ❌ LLMs got much, much better. So the new Browser Use gives the model the full browser 🌐 > Custom Rust harness + TUI > Direct CDP control > Full browser action space > Self-recovers from failures For the first time, we trust the agent to finish the job ✅ 0.13.0 - same interface, completely new architecture. https://lnkd.in/gW5yp4MJ

Introducing: New Browser Infrastructure. ⚡ Chromium fork. Firecracker fork. Custom Linux kernel. > 3x cheaper. $0.02/h. > Unlimited scaling > Subsecond cold starts > Most stealth browser infra Feel the speed. Now live. Try it here: https://lnkd.in/gyy5UnZn

Introducing: Free Tier for Browser Use Cloud 🚀 We’re giving all agents their own cloud browsers! > Unlimited browser hours > Free proxies > Persistent authentication Let your agents try for free at https://browser-use.com/

The best AI researcher isn't a human anymore. We beat OpenAI, Google, and Anthropic on the biggest browser agent benchmark. Our researcher is an LLM + CLI that:  → Spots its own failures  → Finds the root cause  → Fixes and loops - forever Watch it and tell me I'm wrong.