Brow

What it does

Brow is an interactive AI agent that lives in the Chrome side-panel, allowing users to automate browser tasks using a combination of local LLMs and MCP tools. Unlike detached automation browsers, Brow operates within a live user session, providing a seamless interface for semantic browsing.

Key features

• Browser Snapshots: Uses compact, ref-based snapshots of the live DOM for efficient reasoning and interaction.
• WebMCP Integration: Dynamically discovers and uses page-local tools exposed via navigator.modelContext on active tabs.
• Remote MCP Support: Connects to remote MCP servers via HTTP/SSE to bring external tool capabilities into the browser.
• Workflow Memory: Records and stores reusable skills and domain memory to improve automation over time.
• Semantic Forms: Provides structured snapshots of form fields for batch processing and automated data entry.

Installation

• Local Build:
  1. Clone the repository.
  2. Run npm install then npm run build.
  3. Load the dist/ folder as an unpacked extension in chrome://extensions (Developer mode).
• Configuration: Set your API keys and model endpoints in the Brow Options page.

Supported hosts

• Chrome (via Extension)