BrowserSkill enables AI agents to interact with real, logged-in browser sessions without disrupting user workflows. It targets AI agents like Cursor, Claude Code, and Codex, allowing them to leverage existing authenticated web states, thereby eliminating the need for separate test accounts and providing a seamless user experience.

How It Works

BrowserSkill functions as a local bridge between AI agent harnesses and the user's browser. It comprises a bsk CLI/daemon and a browser extension. Agents issue commands to the bsk CLI, which are relayed to a local daemon. This daemon then communicates with the browser extension via WebSocket, directing it to execute tasks within a dedicated "Agent Window." This architecture ensures agents never directly access the user's primary browser windows, preserving user privacy and workflow continuity. The approach is advantageous for its agent agnosticism, leveraging shell capabilities for broad compatibility.

Quick Start & Requirements

• Primary Install/Run: For users of supported agents (Cursor, Claude Code, etc.), the recommended installation is a single command provided to the agent: Set up browser-skill on this machine by following https://raw.githubusercontent.com/Tencent/BrowserSkill/main/AGENT_INSTALL.md.
• Manual Install:
  1. Install the bsk CLI: macOS/Linux: curl -fsSL https://raw.githubusercontent.com/Tencent/BrowserSkill/main/install.sh | sh; Windows: irm https://raw.githubusercontent.com/Tencent/BrowserSkill/main/install.ps1 | iex.
  2. Install the browser extension from the Chrome Web Store.
  3. Install the skill for your agent harness: bsk install-skill.
• Prerequisites:
  • Operating Systems: macOS (Apple Silicon/Intel), Linux (x64/ARM64), Windows x64.
  • Browsers: Chrome and Microsoft Edge are supported. Other Chromium-based browsers are expected to work. Firefox support is planned.
• Links:
  • Agent Install Guide: https://raw.githubusercontent.com/Tencent/BrowserSkill/main/AGENT_INSTALL.md
  • Chrome Extension: https://chromewebstore.google.com/detail/hhcmgoofomhgciiibhipgmgkgnoenaoi

Highlighted Details

• Reuses real login states for agents, eliminating the need for separate test accounts.
• Agent tasks run in a dedicated "Agent Window," allowing users to continue working uninterrupted in their main browser windows.
• Supports any shell-capable AI agent via the bsk CLI, avoiding vendor lock-in.
• Includes a built-in human-in-loop mechanism for handling captchas, logins, or other human-only steps, allowing agents to request user intervention.

Maintenance & Community

The repository is structured as a Cargo + pnpm workspace. No specific community channels (e.g., Discord, Slack) or detailed contributor information are provided in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: The permissive MIT license suggests broad compatibility for commercial use and integration into closed-source projects.

Limitations & Caveats

Currently, official support is limited to Chrome and Edge browsers, with Firefox support listed as planned. The README does not detail specific performance benchmarks or known bugs.