OpenCLI: a framework to convert websites and browser sessions into deterministic CLI interfaces for humans and AI agents

What it solves

OpenCLI converts websites, browser sessions, and Electron apps into deterministic command-line interfaces (CLIs). It allows humans and AI agents to interact with web-based services—navigating, filling forms, and extracting data—without needing to manually operate a browser UI, while leveraging the user's existing logged-in sessions.

How it works

OpenCLI uses a lightweight Browser Bridge extension and a local daemon to connect to Chrome/Chromium. It provides a set of primitives (like open, click, type, and extract) that can be triggered via the command line. For AI agents, it provides specific "skills" (e.g., opencli-browser) that allow LLMs to drive the browser ad-hoc using structured DOM snapshots rather than screenshots.

Who it’s for

• Developers and Power Users who want to automate web tasks or access site data via a terminal.
• AI Agent Developers who want to give their agents the ability to operate real websites using a user's authenticated browser session.
• Automation Engineers looking to create reusable adapters for specific websites.

Highlights

• AI-Ready Skills: Dedicated skills for agents (like Claude Code or Cursor) to navigate and interact with the web.
• Extensive Built-in Adapters: Pre-configured commands for over 100 sites, including X (Twitter), Reddit, Bilibili, and LinkedIn.
• CLI Hub: Acts as a unified interface for other local binaries (e.g., gh, docker) and Electron apps (e.g., Cursor, ChatGPT App).
• Multi-Profile Support: Ability to manage and switch between different Chrome profiles.
• Flexible Output: Supports multiple data formats including JSON, YAML, CSV, and Markdown.