jackwener/OpenCLI

What it solves

OpenCLI transforms websites into deterministic command-line interfaces (CLIs) and provides a way for AI agents to interact with the web using a user's existing, logged-in Chrome sessions. It eliminates the need for site-specific adapters for every task by allowing agents to navigate, click, and extract data directly from any page, while also providing a library of pre-built adapters for popular platforms.

How it works

OpenCLI uses a combination of a local daemon and a "Browser Bridge" Chrome extension to connect a CLI runtime to a live browser. This allows it to:

• Execute Pre-built Adapters: Run specific commands for sites like Reddit, X (Twitter), and LinkedIn to extract data or perform actions.
• Enable AI Agent Skills: Provide a set of "skills" (via npx skills) that AI agents (like Claude Code or Cursor) can use to drive the browser, read DOM snapshots, fill forms, and intercept network responses.
• Author New Adapters: Use a specialized skill (opencli-adapter-author) to analyze a site's structure and generate new reusable CLI commands.

Who it’s for

• AI Agent Developers: Who need their agents to operate real websites using authenticated user sessions.
• Power Users: Who want to automate web tasks or scrape data from various platforms via the terminal without writing custom scripts for every site.
• Developers: Who want to quickly create and share reusable browser automation plugins.

Highlights

• Logged-in Session Access: Operates within the user's active Chrome profile, bypassing many authentication hurdles.
• Agent-Ready: Includes dedicated skills for AI agents to navigate, interact, and extract data ad-hoc.
• Extensive Built-in Library: Supports over 100 sites including Bilibili, Zhihu, and Amazon.
• Deterministic Output: Supports multiple output formats including JSON, YAML, CSV, and Markdown for easy piping into other tools or LLMs.
• Plugin System: Allows users to install and share community-contributed adapters via GitHub.