Nanobrowser is an open-source Chrome extension for AI-powered web automation, offering a privacy-focused, flexible alternative to paid services like OpenAI Operator. It enables users to build and run multi-agent workflows directly within their browser, leveraging their own LLM API keys for cost control and data security.

How It Works

Nanobrowser employs a multi-agent system architecture, typically comprising a Planner, Navigator, and Validator. The Planner breaks down complex tasks, the Navigator executes actions within the browser, and the Validator assesses task completion. This modular design allows for specialized LLM assignments to each agent, optimizing for performance and cost. The extension operates locally, ensuring user data and credentials remain private.

Quick Start & Requirements

• Install: Via Chrome Web Store or manually by downloading the latest nanobrowser.zip from releases.
• Prerequisites: Node.js (v22.12.0+), pnpm (v9.15.1+) for building from source. LLM API keys (OpenAI, Anthropic, Gemini, Ollama, or OpenAI-compatible) are required for functionality.
• Setup: Installation is quick via Chrome Web Store. Configuring LLM API keys takes minutes.
• Docs: GitHub Releases

Highlighted Details

• Supports multiple LLM providers including OpenAI, Anthropic, Gemini, and Ollama.
• Enables flexible LLM configuration per agent (e.g., Claude 3.7 Sonnet for planning, Claude 3.5 Haiku for navigation).
• Allows running local LLMs via Ollama for zero API costs and enhanced privacy.
• Features an interactive side panel for chat and real-time status updates.

Maintenance & Community

• Active development with a roadmap available on GitHub Discussions.
• Community channels include Discord and Twitter.
• Contributions are welcomed via code, feedback, and prompt sharing.

Licensing & Compatibility

• Licensed under Apache License 2.0.
• Permits commercial use and integration with closed-source projects.

Limitations & Caveats

The Chrome Web Store version may lag behind the latest features due to the review process. Performance and stability with cost-effective or local LLM configurations may require more prompt engineering and iteration for complex tasks.