BrowserOS - Open-Source Agentic Browser With AI Agents

TL;DR

TL;DR: BrowserOS is an open-source Chromium fork with native AI agents, MCP server support, scheduled tasks, and 40+ app integrations — a privacy-first alternative to ChatGPT Atlas, Perplexity Comet, and Dia.

Source and Accuracy Notes

• Project page: BrowserOS.com
• Source repository: github.com/browseros-ai/BrowserOS
• License: AGPL-3.0 (verified via GitHub API license.spdx_id)
• Documentation: docs.browseros.com
• HN launch thread: news.ycombinator.com/item?id=47170174
• Source last checked: 2026-06-16 (latest release v0.44.0.1)

What Is BrowserOS?

BrowserOS is an open-source Chromium-based browser with AI agents built directly into the browsing experience. Instead of installing separate extensions or running terminal-based automation tools, you describe tasks in plain language and the browser handles clicking, typing, and navigating on your behalf.

The project positions itself as a privacy-first alternative to closed-source AI browsers like ChatGPT Atlas, Perplexity Comet, and Dia. All AI processing can run locally via Ollama or LM Studio, or you can bring your own API keys for Claude, OpenAI, Gemini, and other providers.

Key numbers from the project:

• 11,000+ GitHub stars
• 53 built-in browser automation tools
• 40+ app integrations via MCP (Gmail, Slack, Notion, GitHub, Linear, Figma, Salesforce)
• Supports macOS, Windows, and Linux

Why BrowserOS Matters

The AI browser space is heating up. ChatGPT Atlas, Perplexity Comet, and Dia all launched with proprietary, cloud-dependent architectures. BrowserOS takes the opposite approach:

1. Open source (AGPL-3.0) — you can audit, fork, and self-host
2. Local-first privacy — run models on your own hardware, your data never leaves your machine
3. Bring your own keys — use existing ChatGPT Pro, GitHub Copilot, or Claude subscriptions via OAuth instead of paying twice
4. MCP server built in — control the browser from Claude Code, Gemini CLI, or any MCP-compatible client

Setup Workflow

Step 1: Download and Install

BrowserOS provides native installers for all major platforms:

# macOS
# Download from: https://files.browseros.com/download/BrowserOS.dmg

# Windows
# Download from: https://files.browseros.com/download/BrowserOS_installer.exe

# Linux (AppImage)
# Download from: https://files.browseros.com/download/BrowserOS.AppImage

# Linux (Debian/Ubuntu)
# Download from: https://cdn.browseros.com/download/BrowserOS.deb

After installation, you can optionally import your Chrome data — bookmarks, passwords, and extensions all carry over.

Step 2: Connect Your AI Provider

BrowserOS supports multiple authentication methods:

Cloud providers (API key):
  - Claude (Anthropic)
  - GPT-4o / o3 (OpenAI)
  - Gemini (Google)
  - Azure OpenAI
  - AWS Bedrock
  - OpenRouter

Cloud providers (OAuth — use existing subscriptions):
  - ChatGPT Pro/Plus
  - GitHub Copilot
  - Qwen Code

Local models:
  - Ollama
  - LM Studio

Default provider:
  - Kimi K2.5 (built-in, no setup required)

Step 3: Install the CLI (Optional)

The browseros-cli lets you control BrowserOS from the terminal or from AI coding agents:

# macOS / Linux
curl -fsSL https://cdn.browseros.com/cli/install.sh | bash

# Windows (PowerShell)
irm https://cdn.browseros.com/cli/install.ps1 | iex

# Initialize connection to running BrowserOS instance
browseros-cli init

Step 4: Use as MCP Server

BrowserOS ships with a built-in MCP server. Connect it to Claude Code, Gemini CLI, or any MCP client to automate browser tasks from your terminal workflow.

# In Claude Code or Gemini CLI, add BrowserOS as an MCP server
# The browser exposes 53+ tools for navigation, clicking,
# typing, data extraction, and more

See the MCP setup guide for configuration details.

Deeper Analysis

Architecture

BrowserOS is a monorepo with two main subsystems:

BrowserOS/
├── packages/browseros/              # Chromium fork + build system (Python)
│   ├── chromium_patches/            # Patches applied to Chromium source
│   ├── build/                       # Build CLI and modules
│   └── resources/                   # Icons, entitlements, signing
│
├── packages/browseros-agent/        # Agent platform (TypeScript/Go)
│   ├── apps/
│   │   ├── server/                  # MCP server + AI agent loop (Bun)
│   │   ├── agent/                   # Browser extension UI (WXT + React)
│   │   ├── cli/                     # CLI tool (Go)
│   │   ├── eval/                    # Benchmark framework
│   │   └── controller-ext/          # Chrome API bridge extension
│   │
│   └── packages/
│       ├── agent-sdk/               # Node.js SDK (npm: @browseros-ai/agent-sdk)
│       ├── cdp-protocol/            # CDP type bindings
│       └── shared/                  # Shared constants

The agent server runs on Bun and exposes 53+ MCP tools. The browser extension uses WXT + React for the UI. The CLI is written in Go.

Key Differentiators vs Chrome/Brave

| Feature | BrowserOS | Chrome | Brave | |---------|-----------|--------|-------| | Open Source | Yes | No | Yes | | AI Agent | Yes | No | No | | MCP Server | Yes | No | No | | Visual Workflows | Yes | No | No | | Scheduled Tasks | Yes | No | No | | Bring Your Own Keys | Yes | No | Yes | | Local Models (Ollama) | Yes | No | Yes | | Ad Blocking (MV2) | Yes | No | Yes |

Feature Highlights

• Cowork — combine browser automation with local file operations. Research the web, save reports directly to your filesystem.
• Scheduled Tasks — run agents on autopilot: daily, hourly, or every few minutes.
• Workflows — build repeatable browser automations with a visual graph builder.
• Memory — persistent context across conversations.
• SOUL.md — define your AI’s personality and instructions in a single markdown file.
• LLM Hub — compare Claude, ChatGPT, and Gemini responses side-by-side on any page.
• Ad Blocking — uBlock Origin with Manifest V2 support (10x more protection than Chrome per the project docs).

Practical Evaluation Checklist

Before adopting BrowserOS, evaluate:

• [ ] LLM cost — do you already have a ChatGPT Pro or GitHub Copilot subscription? If yes, BrowserOS can use it via OAuth at no extra cost.
• [ ] Privacy requirements — does your workflow require local-only inference? BrowserOS supports Ollama and LM Studio for fully offline operation.
• [ ] MCP ecosystem — do you use Claude Code, Gemini CLI, or other MCP clients? BrowserOS’s built-in MCP server integrates directly.
• [ ] Automation needs — do you need scheduled, recurring browser tasks? The built-in scheduler handles this without external tools like n8n or Selenium.
• [ ] Platform support — macOS, Windows, and Linux are all supported with native installers.
• [ ] Extension compatibility — BrowserOS is a Chromium fork, so most Chrome extensions work. Manifest V2 extensions (like uBlock Origin) are fully supported.

Security Notes

• AGPL-3.0 license — source code is fully auditable. The AGPL requires that any modified version deployed over a network must also be open-sourced.
• Local-first option — when using Ollama or LM Studio, no data leaves your machine. Cloud providers send prompts to their respective APIs.
• OAuth for existing subscriptions — BrowserOS uses OAuth to authenticate with ChatGPT Pro and GitHub Copilot. Your credentials are handled by the provider’s auth flow, not stored by BrowserOS.
• Sandboxed agent execution — the AI agent runs in a sandboxed environment within the browser, with explicit tool permissions.
• ungoogled-chromium patches — BrowserOS incorporates privacy patches from the ungoogled-chromium project for enhanced tracking protection.

FAQ

Q: Is BrowserOS free to use? A: Yes, BrowserOS is free and open-source (AGPL-3.0). You only pay for the LLM provider you choose. If you use local models via Ollama, there are zero ongoing costs. If you have a ChatGPT Pro or GitHub Copilot subscription, you can use it via OAuth at no extra cost.

Q: Does BrowserOS work with Chrome extensions? A: Yes. BrowserOS is a Chromium fork, so most Chrome extensions are compatible. It also supports Manifest V2 extensions, which means uBlock Origin works with full filtering capabilities (unlike Chrome, which is phasing out MV2).

Q: Can I use BrowserOS without an internet connection? A: Yes, when paired with a local model provider like Ollama or LM Studio. The browser itself works offline for local file operations, and the AI agent can run fully locally.

Q: How does BrowserOS compare to using Playwright or Selenium for automation? A: BrowserOS is designed for natural-language automation — you describe what you want in plain English. Playwright and Selenium require writing code. BrowserOS is better for ad-hoc tasks and non-developers; Playwright/Selenium are better for deterministic, test-suite automation.

Q: What AI models are supported? A: BrowserOS supports Claude, GPT-4o, o3, Gemini, Azure OpenAI, AWS Bedrock, OpenRouter, Kimi K2.5 (default), ChatGPT Pro/Plus (via OAuth), GitHub Copilot (via OAuth), Qwen Code (via OAuth), and any model available through Ollama or LM Studio locally.

Q: Is my data sent to third parties? A: Only if you choose a cloud LLM provider. When using local models (Ollama, LM Studio), all processing stays on your machine. The project states it is “privacy-first” and does not collect browsing data.

Conclusion

BrowserOS fills a gap in the AI browser market by combining open-source transparency with native AI agent capabilities. Its MCP server integration makes it uniquely suited for developers who already use Claude Code or Gemini CLI, while the OAuth support for existing subscriptions (ChatGPT Pro, GitHub Copilot) means you can get AI-powered browser automation without additional cost.

The project is actively developed (v0.44.0 released March 2026) and has gained significant traction with over 11,000 GitHub stars since its May 2025 launch. If you want an AI browser that does not lock you into a proprietary ecosystem, BrowserOS is worth evaluating.