BrowserOS - Open-Source AI Browser for Agents

TL;DR

TL;DR: BrowserOS is a local-first, open-source Chromium browser with built-in AI for humans; BrowserClaw gives AI agents (Claude Code, Codex, Cursor) real browser control — solving the “AI can’t press the buttons” problem.

What Is BrowserOS?

BrowserOS is an open-source Chromium fork with two products in one codebase:

• BrowserOS — the AI browser for humans, a privacy-first alternative to ChatGPT Atlas, Perplexity Comet, and Dia
• BrowserClaw — the browser for AI agents, letting tools like Claude Code, Codex, and Cursor drive a real browser using accounts you’re already signed into

Both are free, open-source, local-only, and require you to bring your own AI keys. The project is AGPL-3.0 licensed and has 12,416 GitHub stars as of July 2026.

Why This Matters

AI assistants are powerful, but they hit a wall when a task requires clicking a button, filling a form, or accessing a logged-in account. BrowserClaw solves this by giving your AI agent its own Chromium instance it can actually control.

Setup Workflow

Step 1: Install BrowserClaw

Download from the official site for your platform:

• macOS: BrowserClaw.dmg
• Windows: BrowserClaw_installer.exe

Install like any browser, then sign into the sites you use daily.

Step 2: Connect Your AI

BrowserClaw ships with one-click setup for seven AI tools:

• Claude Code
• Codex
• Cursor
• VS Code
• Zed
• OpenCode
• Antigravity

For anything else that speaks MCP, connect with a URL endpoint.

Step 3: Give Your Agent a Real Task

In your AI chat:

“Find a good time next week for a 30-minute team meeting and send the invite.”

BrowserClaw opens its own tab, navigates to your calendar, and does the work while you watch live. Every session is saved as a replayable video.

Key Features

Live Dashboard

BrowserClaw’s new tab shows every active agent session — which site it’s on, what it’s doing, and how far along. You can monitor multiple agents simultaneously.

Session Replay

Every session is saved as a scrubbable video on your local disk, complete with a step-by-step action timeline. Rewind and inspect exactly what the agent did.

Privacy-First Architecture

All browsing runs locally. No data leaves your machine. You bring your own AI API keys. The default branch is main and the project is actively maintained (last push: July 20, 2026).

System Requirements

| Component | Requirement | |---|---| | OS | macOS, Windows, Linux (AppImage), Debian | | AI integration | MCP-compatible client | | AI keys | User-provided (no built-in billing) | | License | AGPL-3.0 |

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: BrowserOS.com ← verified July 20, 2026
• Source repository: github.com/browseros-ai/BrowserOS ← README read July 20, 2026
• License: AGPL-3.0 ← verified via GitHub API license.spdx_id
• GitHub stars: 12,416 (verified via GitHub REST API)
• HN launch thread: news.ycombinator.com/item?id=46721474 ← Show HN: BrowserOS – “Claude Cowork” in the browser
• Source last checked: 2026-07-20 (commit main branch)

FAQ

Q: What is the difference between BrowserOS and BrowserClaw? A: BrowserOS is an AI-powered browser for humans (like a privacy-focused alternative to AI browsers such as ChatGPT Atlas). BrowserClaw is the same codebase configured for AI agents — giving tools like Claude Code control of a real Chromium browser signed into your accounts.

Q: Does BrowserOS use my own AI API keys? A: Yes. You bring your own keys. BrowserOS does not include built-in AI billing or a hosted model.

Q: Is all browsing local? A: Yes. All sessions run locally on your machine. The project is open-source and the data stays on-disk.

Q: Which AI tools can connect to BrowserClaw? A: BrowserClaw ships with one-click MCP setup for Claude Code, Codex, Cursor, VS Code, Zed, OpenCode, and Antigravity. Any MCP-compatible client can connect via URL.

Q: What license is BrowserOS released under? A: AGPL-3.0.

Conclusion

BrowserOS solves a real gap in the AI agent ecosystem: the inability to interact with the live web through a real browser session. BrowserClaw turns any MCP-compatible AI tool into a browser-capable agent, with full live monitoring and session replay. It’s actively maintained, fully open-source, and runs entirely locally — no data leaves your machine.

If you run AI coding agents in your workflow, BrowserClaw is worth adding to your stack.