AI Native Landscape

OpenCLI

Tracked

A browser automation CLI for AI agents that converts websites into deterministic CLI commands and lets agents operate any website through your logged-in Chrome via a bridge extension, with reusable site adapters.

AuthorjackwenerOpen Sourced2026-03-14Last CommitUnknown

Overview

OpenCLI turns websites and browser sessions into deterministic interfaces for humans and AI agents. It is a general-purpose browser automation CLI that connects to your real, logged-in Chrome through a Browser Bridge extension plus a local daemon, so commands run with your existing sessions — no site adapter required for arbitrary pages. Built-in adapters add reusable commands for sites like Bilibili, Zhihu, Xiaohongshu, Reddit, HackerNews, and Twitter/X, and an authoring workflow lets you write new adapters end-to-end.

Key Features

  • One surface, three uses: run built-in site adapters, let AI agents operate any website through opencli browser primitives, or author new adapters with the opencli-adapter-author skill
  • Agent command contract: JSON on stdout ({"ok": true, "data": ...} or {"ok": false, "error": {code, message, hint}}), logs on stderr, nonzero exit on failure
  • Works with your logged-in browser — no headless browser or separate automation profile
  • Built-in adapters for Bilibili, Zhihu, Xiaohongshu, Reddit, HackerNews, Twitter/X, and more
  • Skills for AI agents (opencli-browser for Claude Code, Cursor, etc.) covering navigate, click, type/fill, extract, and inspect
  • Multi-profile support: list connected Chrome profiles and assign local aliases
  • OpenCLIApp desktop bundles the runtime with a tray UI for setup, diagnostics, updates, and login keepalive

Use Cases

  • Giving AI agents reliable, deterministic commands against websites instead of fragile screenshot-driven automation
  • Reading and acting on sites behind logins (social platforms, internal tools) using your existing sessions
  • Building reusable site adapters for repeated workflows like search, listing, and extraction
  • Humans running quick site queries (opencli hackernews top --limit 5) without opening a browser

Technical Details

  • Node.js CLI (>= 20.18.1, npm @jackwener/opencli) that drives Chrome/Chromium through a lightweight Browser Bridge extension plus a local daemon that auto-starts on demand
  • Deterministic output contract for agents: commands default to JSON {ok, data}/{ok, error} on stdout with structured error codes, messages, and hints; browser follow modes stream JSON Lines
  • Adapter system: built-in site adapters plus an authoring pipeline (opencli browser primitives + opencli-adapter-author skill + adapter init|check|test) that guides from first recon through field decoding to tested code
  • Page scoping via --session isolates browser operations; native Chrome calls bypass sessions; profiles map Chrome contextIds to local aliases with an explicit choice when multiple are connected
  • opencli doctor verifies the setup; state --diagnostics and network --static replace older analyze/discover flows