BrowserOS – The Open Source Agentic Browser
Build AI agents that automate any browser task using plain English
TL;DR: BrowserOS is an open-source, AI-native browser that automates your tedious web tasks. Just describe what you want in plain English—like "find software engineers from my LinkedIn requests and add them to a google sheet"—and watch an AI click, type, and navigate on your behalf, all running locally with your logged-in sessions. We're an open-source and privacy-first alternative to Comet, Dia, Chrome.
We are grateful for all our early supporters — 4.3k+ Github ⭐, 25,000+ downloads and an active discord community of 1000 folks!
🤔 Why Now
AI coding tools made developers 10x faster. But a billion knowledge workers still grind through browser busywork. Copy-paste. Click. Repeat. Even we spend half our engineering time on mindless browser tasks. It's ridiculous—AI writes our entire codebase but can't click a button.
So what’s blocking the browser’s ‘Cursor moment’?
🧨 The Problem
We think it comes down to three big problems preventing mass adoption of AI agents for knowledge workers, especially at enterprises:
1. Can't access your logged-in sessions. Most agentic solutions today (like ChatGPT Operator, Manus) run on remote computers. They can't log into your Gmail, LinkedIn, or company tools. So they fail at every real task you actually need help with.
2. Tools are fragmented. You've got agent builders that use MCP servers. Others that only do browser automation. Some that just call APIs like Zapier. There's no single solution that lets you build an agent using all these approaches together.
3. Closed source: Most other popular browsers come from search/ad companies who want to track you and sell ads. They lock you into their LLMs, hide how things work, and send your data to their servers. That's a dealbreaker for enterprises who can't send sensitive data to third parties.
🥳 Our Solution
BrowserOS is an open-source, privacy-first browser built to solve these problems by design.
- It's just a normal browser you download and run on your machine. You log into websites like you always do. But now agents can use those logged-in sessions to actually do useful work - they can browse the web just like you would, using your existing logins.
- We support both MCP servers and browser automation. You can stitch them together to build agents that handle complex workflows. Need to pull data from a website, process it through an API, then update a spreadsheet? One agent can do it all.
- We're fully open source. You can see our system prompts, change them if you want. Bring your own API keys or run models locally with Ollama. Your data stays on your computer.
Key features of BrowserOS:
- Build agents with plain English - Just describe what you want. No coding required.
- Use any LLM you want - Bring your own API keys, switch between models, or run everything using local LLMs
- Still a normal browser - It's Chromium underneath. All your Chrome extensions work.
🚀 The future
Here's what excites us: a billion knowledge workers spend 60-80% of their day in browsers. Every enterprise runs on web apps—Salesforce, SAP, Workday, internal tools. These workers waste hours on tasks that make them miserable. Filing expense reports. Extracting data from dashboards. Setting up ad campaigns.
We think browsers will become new operating system where AI employees live and work alongside humans. Imagine a world where IT builds an “expense report agent” once, every employee benefits. Someone shares their “Facebook Ads agent” publicly. You fork it, tweak it for your workflow. The mundane work gets automated while humans focus on what matters.
This kind of automation only works in browsers. Not everything has APIs, but everything has a web UI. If an agent can click and type like a human, it can automate any web app—from legacy enterprise software to the latest SaaS tools.
We're building the best browser for this future!
🏋️♂️ Team
We're twin brothers who've been building products together for years. We spent the last 6 years as engineers at Google and Meta.
Nikhil worked on Instagram Reels and Facebook Feed Ranking backend—deep C++ and systems work. Nithin was an ML engineer on YouTube Recommendations and worked on building Youtube’s first Large Recommender Model (LRM).
We feel this gives us the perfect mix of experience to tackle a massive C++ codebase like Chromium while building a powerful agent and AI layer on top.
