nanobrowser/nanobrowser
Nanobrowser is an open-source Chrome extension that brings AI-powered web automation directly into the browser using a multi-agent architecture. It serves as a free, self-hosted alternative to services like OpenAI Operator — users bring their own LLM API keys and keep full control over their data and costs. Two specialized agents collaborate on each task: a Planner agent that breaks down the goal into steps, and a Navigator agent that executes those steps by interacting with web pages. Both agents can be assigned different models from providers including OpenAI, Anthropic (Claude), Google Gemini, Ollama, Groq, and custom endpoints. The extension adds a real-time chat sidebar to Chrome where users describe tasks in natural language — extracting headlines, searching products, researching GitHub repositories, filling forms, or any multi-step web workflow. Conversation history is preserved for review. Installation is available via the Chrome Web Store or by loading the unpacked extension from a release ZIP in developer mode. No server infrastructure is required; everything runs locally through the extension with direct API calls to the chosen LLM provider. The recommended configuration is Claude Sonnet as Planner and Claude Haiku as Navigator for the best cost-to-performance ratio.
git clone https://github.com/nanobrowser/nanobrowser.git
Quick Start Example
// 1. Install from Chrome Web Store or load unpacked
// 2. Open sidebar, go to Settings → add your API key
// Example tasks you can type in the chat:
// "Go to github.com/trending and extract the top 10 repos today"
// "Search Amazon for mechanical keyboards under and list specs"
// "Open Hacker News and summarize the top 5 stories"
// Recommended model setup:
// Planner → claude-sonnet-4-5
// Navigator → claude-haiku-4-5