browseros-ai/BrowserOS

🌐 The open-source Agentic browser; alternative to ChatGPT Atlas, Perplexity Comet, Dia.

BrowserOS – An Agent‑friendly Chromium fork

What it is – BrowserOS is an open‑source Chromium‑based browser that comes in two flavors:

  • BrowserOS neo – a secondary browser that is driven by AI agents. It lets you import your Chrome logins, connect agents such as Claude Code, Codex, Cursor, VS Code, etc., and hand them web‑tasks (e.g., “book a flight”). Each agent runs in its own tab, you can watch the live progress on a dashboard, and later replay the whole session as a scrubbable video.
  • BrowserOS – the opposite side: a regular browser that you drive, with an AI assistant built into every new tab. It can use your own LLM keys or a local Ollama model.

Both browsers run locally (macOS, Windows, Linux) and keep all data – logins, screenshots, session recordings – on your machine, never sending page content to the cloud.

Key technical pieces

  • Chromium fork – located in packages/browseros/ (C++, Python patches). Uses ungoogled‑chromium patches for extra privacy.
  • Agent platform – TypeScript + Rust code in packages/browseros-agent/. It provides:
    • MCP server (claw-server) – a multi‑client protocol endpoint that agents talk to.
    • Dashboard UI (claw-app) – a React/WXT extension showing live agent tabs, progress, and replay controls.
    • Bun‑based backend (server) – exposes the MCP tools and runs the AI‑agent loop.
    • Rust crates for CDP bindings, core primitives, and harness integrations.
  • One‑click integration – the UI can discover and connect to supported “harnesses” (Claude Code, Codex, Cursor, VS Code, OpenClaw, Hermes, etc.) without manual configuration.
  • Replay & audit – every agent session is saved as a video plus a step‑by‑step action timeline under ~/.browserclaw/.

Why it matters

  • Existing headless drivers (Playwright, Selenium) start a fresh browser with no logged‑in state, which makes many real‑world automation tasks impossible. BrowserOS imports your existing Chrome credentials, so agents can act as you.
  • Cloud‑browser services run in datacenters, often get blocked by sites and cannot use your personal accounts. BrowserOS runs locally, avoiding IP‑based blocks and preserving privacy.
  • It is not locked to a single LLM vendor; any tool that implements the MCP (Multi‑Client Protocol) can be hooked up.

Getting started

brew tap browseros-ai/tap
brew install --cask browseros-neo   # macOS
# or download the installer for Windows / Linux from the badges above

Then import Chrome data, connect an agent, and issue a natural‑language task.

Community & contribution

  • Discord, Slack, and a public X account for support.
  • Contributing guides for the TypeScript/Rust side (≈15 min) and for the Chromium side (requires ~100 GB disk space).
  • Licensed under AGPL‑3.0.

Bottom line – BrowserOS provides a full‑featured, privacy‑first Chromium browser that can be programmatically controlled by your favourite AI coding agents, enabling real‑world web automation while keeping everything on‑device.

Related

  • Project
    anomalyco/browser-control

    Browser Control is an open‑source Node.js driver that lets AI agents or CLI users run Playwright code inside their own Chromium browser (with all logged‑in sessions and extensions). It works via a local relay and a browser extension, supports persistent sessions, inspection helpers, human hand‑offs, video/ HAR recording, and a TypeScript Effect client for secret‑safe API calls. Intended for trusted local use only.

  • Project
    platonai/Browser4

    Browser4 is an AI‑native browser automation engine that lets LLM agents (or humans) drive real browsers via a fast Rust CLI, extract data with zero‑token X‑SQL/CSS selectors, and run large‑scale crawls or ML‑based spreadsheet generation (WebMiner). It combines CDP‑native performance, swarm scaling, and hybrid LLM/ML intelligence, and is distributed as an npm CLI with a Java backend (Maven Central).

  • Project
    iFurySt/open-browser-use

    Open Browser Use is an open‑source, runtime‑agnostic browser‑automation layer for AI agents. It pairs a Chrome extension with a CLI and provides JavaScript, Python, and Go SDKs plus an MCP server, enabling agents to list tabs, open URLs, navigate, and run CDP actions. Install via npm or Homebrew, run `open-browser-use setup` to register the native host, then add the skill to agents like Codex or Claude‑Code. MIT‑licensed.

  • Project
    browser-use/browser-use

    Browser Use is an open‑source Python/TypeScript library that gives LLM agents full control over a real or cloud‑hosted web browser. It provides 31 browser actions as tools, supports custom functions, works with OpenAI, Anthropic, Google models or its own BU2 model, and offers three usage modes: a fully hosted cloud API, a CLI for existing agents, and a library you embed in your code. Ideal for automating web tasks, building AI assistants that browse, or researching LLM‑driven web navigation.