Open-source Chrome extension with multi-agent AI for web automation. Free alternative to OpenAI Operator. 12K+ stars.
TO
TokRepo精选 · Community
Quick Use
Use it first, then decide how deep to go
This block should tell both the user and the agent what to copy, install, and apply first.
1. Install from [Chrome Web Store](https://chromewebstore.google.com/detail/nanobrowser/imbddededgmcgfhfpcjmijokokekbkal) or download from GitHub Releases
2. Click the Nanobrowser icon in Chrome toolbar
3. Enter your LLM API key (OpenAI, Anthropic, or Gemini)
4. Type a task in natural language:
```
Find the cheapest flight from NYC to London next Friday on Google Flights
```
5. Watch the AI agents navigate, click, and extract data in real-time
For local/custom models, configure an OpenAI-compatible endpoint (Ollama, vLLM, etc.) in Settings.
---
Intro
Nanobrowser is an open-source Chrome extension with 12,600+ GitHub stars that brings AI-powered web automation directly into your browser. Positioned as a free alternative to OpenAI Operator, it uses a multi-agent architecture where a Planner agent breaks down tasks and a Navigator agent executes browser actions with intelligent self-correction. Everything runs locally in your browser using your own API keys — no subscription fees, no data sent to third-party servers. Supports OpenAI, Anthropic, Gemini, Ollama, Groq, and any OpenAI-compatible endpoint.
Works with: Chrome, Edge, OpenAI GPT-4o, Anthropic Claude, Google Gemini, Ollama, Groq, Cerebras. Best for users who want AI browser automation without coding or subscriptions. Setup time: under 2 minutes.
---
## Nanobrowser Multi-Agent Architecture
### Agent Roles
```
┌──────────────────────────────┐
│ User Request │
│ "Book a table for 2 at..." │
└──────────┬───────────────────┘
│
┌──────▼──────┐
│ Planner │ Breaks task into steps
│ Agent │ Monitors progress
└──────┬──────┘ Self-corrects on errors
│
┌──────▼──────┐
│ Navigator │ Executes browser actions
│ Agent │ Clicks, types, scrolls
└──────┬──────┘ Reads page content
│
┌──────▼──────┐
│ Validator │ Checks task completion
│ Agent │ Reports results
└─────────────┘
```
### How It Works
1. **Task Decomposition** — The Planner agent analyzes your request and creates a step-by-step plan
2. **Web Navigation** — The Navigator agent executes actions: clicking buttons, filling forms, scrolling pages
3. **Self-Correction** — When obstacles arise (popups, CAPTCHAs, changed layouts), agents adapt their strategy
4. **Result Delivery** — The Validator confirms task completion and returns results in the side panel
### Flexible Model Assignment
Use different models for different agents to optimize cost and performance:
| Agent | Recommended Model | Why |
|-------|------------------|-----|
| **Planner** | GPT-4o / Claude Sonnet | Needs strong reasoning for task decomposition |
| **Navigator** | GPT-4o-mini / Gemini Flash | Needs fast responses for real-time navigation |
| **Validator** | Any model | Simple verification tasks |
### Supported Providers
| Provider | Models | Setup |
|----------|--------|-------|
| **OpenAI** | GPT-4o, GPT-4o-mini, o1 | API key |
| **Anthropic** | Claude Sonnet, Claude Haiku | API key |
| **Google** | Gemini Pro, Gemini Flash | API key |
| **Ollama** | Llama, Mistral, Qwen (local) | Local endpoint |
| **Groq** | Llama 3, Mixtral (fast inference) | API key |
| **Cerebras** | Llama 3 (ultra-fast) | API key |
| **Custom** | Any OpenAI-compatible API | Base URL + key |
### Privacy & Security
- **100% local execution** — All processing happens in your browser
- **Your keys, your data** — API keys stored locally, never sent to Nanobrowser servers
- **No subscription** — Completely free, pay only for your LLM API usage
- **Open source** — Full code audit available on GitHub
### Example Tasks
```
📋 "Go to Amazon, search for wireless earbuds under $50, and list the top 5 rated ones"
📋 "Open Google Scholar, search for 'transformer architecture', and extract the top 10 paper titles with citation counts"
📋 "Navigate to GitHub trending, filter by Python, and list today's top repositories"
📋 "Go to my Gmail, find the latest email from John, and summarize it"
📋 "Search for 3-bedroom apartments in Austin TX on Zillow under $2000/month"
```
---
## FAQ
**Q: What is Nanobrowser?**
A: Nanobrowser is a free, open-source Chrome extension with 12,600+ GitHub stars that uses AI multi-agent collaboration to automate web tasks directly in your browser. A free alternative to OpenAI Operator.
**Q: Is Nanobrowser safe? Does it send my data anywhere?**
A: Yes, it's safe. Everything runs locally in your browser. Your API keys and browsing data are stored locally and never sent to Nanobrowser's servers. The only external calls are to your configured LLM provider.
**Q: Is Nanobrowser free?**
A: Yes, the extension is 100% free and open-source under Apache-2.0. You only pay for LLM API usage from your chosen provider (OpenAI, Anthropic, etc.).
---
🙏
Source & Thanks
> Created by [Alex Chen](https://github.com/alexchenzl). Licensed under Apache-2.0.
>
> [nanobrowser](https://github.com/nanobrowser/nanobrowser) — ⭐ 12,600+
Thanks to Alex Chen for building a truly free and private AI browser automation tool.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.