<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> WebBrain is an open-source AI browser agent for Chrome and Firefox, enabling users to interact with web pages, automate tasks, and execute multi-step workflows powered by diverse LLMs. It targets users needing advanced browser automation and LLM integration, offering flexibility and powerful task execution.

How It Works

This monorepo provides browser extensions (Chrome MV3, Firefox MV2) integrating with numerous local and cloud LLMs. It employs "Ask" (read-only), "Act" (interactive), and "Dev" (debugging) modes. Core functionality includes page content extraction, browser element interaction, and autonomous agent workflows. Its approach emphasizes LLM provider flexibility, "Plan before Act" for controlled execution, and intelligent context management via auto-compaction.

Quick Start & Requirements

Clone the repository and load the respective src/chrome or src/firefox folder into the browser's extension management page (enable developer mode). A local LLM server (e.g., llama.cpp, Ollama) is recommended, requiring models with at least a 16k token context window for reliable performance. Setup involves configuring the LLM provider URL in extension settings. Documentation is available in the docs/ directory.

Highlighted Details

• Extensive LLM Support: Integrates with over 20 providers, including numerous local options (llama.cpp, Ollama, Jan) and major cloud APIs (OpenAI, Anthropic, Gemini, Groq).
• Multi-Mode Operation: Offers "Ask" (read-only), "Act" (standard interaction), and "Dev" (advanced debugging) modes.
• Advanced Workflow Features: Includes "Plan before Act" for user-approved execution plans, multi-step agent capabilities, and "Smart Context" for managing conversation history within LLM token limits.
• Customization: Supports importing custom skills and exposes tools via a webbrain-tools manifest.

Maintenance & Community

Maintained by Emre Sokullu. Recent updates highlight ongoing development and feature additions. No specific community channels are detailed.

Licensing & Compatibility

Released under the MIT license, permitting commercial use and integration into closed-source projects.

Limitations & Caveats

The Firefox extension is significantly less capable than Chrome's due to the absence of Chrome DevTools Protocol (CDP) access, lacking features like shadow-DOM traversal and CDP-based input simulation. Firefox also faces limitations with temporary add-on persistence and SPA navigation handling. Advanced interactions with Web Components using shadow DOM require specific tools and modes.