OpenCLI: Turn Any Website Into a CLI for Both Humans and AI Agents

In just three weeks, an open-source project went from zero to 16,000 GitHub stars. No press releases. No venture funding. Just a sharp idea executed well.

OpenCLI — the project from developer jackwener — does one thing that sounds almost too good to be true: it turns any website, Electron app, or local binary into a standardized CLI command that both humans and AI agents can use.

No API keys. No tokens burned at runtime. No brittle screen-scraping. Just a deterministic interface — and your existing Chrome login session.

The Problem OpenCLI Solves

Every website has a CLI equivalent somewhere — but getting there is painful. You could reverse-engineer the API, deal with rate limits, write a scraper that breaks next Tuesday, or pay for a third-party service. For AI agents, the options are even worse: browser-use frameworks that burn through context tokens, screenshot-based approaches that hallucinate, or custom integrations that take days to build.

OpenCLI’s insight is simple: you already have a great browser interface. Just expose it as a CLI.

Instead of building yet another API integration, OpenCLI connects to your Chrome session via CDP (Chrome DevTools Protocol), reuses your logged-in state, and gives AI agents a structured, deterministic way to navigate, read, click, type, extract, and verify on any page.

Three Layers of Automation

OpenCLI isn’t just one thing. It offers three different automation surfaces depending on what you need:

1. Built-in Adapters (Zero Config)

OpenCLI ships with 90+ pre-built adapters for sites like Bilibili, Zhihu, Xiaohongshu (Little Red Book), Reddit, HackerNews, Twitter/X, and more. Run opencli list to see everything, then execute with opencli <site> <command>.

These are deterministic — no browser required, no token cost, works every time. Ideal for both human one-liners and AI agent workflows.

2. AI Agent Browser Control (For Any Website)

When a built-in adapter doesn’t exist, OpenCLI’s browser commands become the interface. The opencli-adapter-author skill (installable into Claude Code, Cursor, and other AI agents) guides the agent end-to-end:

• Navigate to any URL using your logged-in Chrome session
• Read page content via structured DOM snapshots (not screenshots)
• Interact — click buttons, fill forms, select options, press keys
• Extract data from the page or intercept network API responses
• Wait for elements, text, or page transitions

The agent handles all the opencli browser commands internally. You just describe what you want in natural language: “Check my Xiaohongshu notifications” or “Write an adapter for douyin trending.”

3. Write Your Own Adapter (For Humans and Agents)

The opencli browser init + verify workflow takes you from site recon through API discovery, field decoding, and adapter writing. OpenCLI persists learned site knowledge to ~/.opencli/sites/<site>/, so the next adapter you write for the same site is faster.

You can also use OpenCLI as a CLI hub — register local binaries (gh, docker, obsidian) and access them through the same discovery surface with opencli register mycli.

Key Numbers

• GitHub Stars: 16,734 (growing)
• Created: March 14, 2026 — less than 6 weeks old
• npm downloads: 52,706 in the last month
• Built-in adapters: 90+
• License: Apache 2.0
• Language: JavaScript
• Prerequisites: Node.js >= 21.0.0 (or Bun >= 1.0)

Comparison: OpenCLI vs. Browser-Use

OpenCLI is frequently compared to browser-use (89K stars), the most popular AI browser automation framework. Here’s how they differ:

The key distinction: OpenCLI is not an AI agent — it’s a runtime for AI agents. It doesn’t make decisions or call LLMs. It provides the browser primitives that AI agents use to operate websites. This makes it lightweight (no token cost), flexible (any AI agent can use it), and composable (mix with existing adapters or custom code).

Desktop App Control: A Surprising Feature

Beyond websites, OpenCLI can control Electron desktop apps via CDP — Cursor, Codex, ChatGPT, Notion, Antigravity, and others. This means your AI agent can drive a full Cursor IDE session from the terminal, using your own logged-in state. No OAuth setup. No API key management. Just opencli browser open cursor://.

This is genuinely novel — I haven’t seen another open-source tool offer this level of desktop app integration through a CLI interface.

Quick Start

# Install
npm install -g @jackwener/opencli

# Install the Chrome extension (required for browser commands)
# Download from https://github.com/jackwener/opencli/releases
# Unzip → chrome://extensions → Load unpacked

# Verify setup
opencli doctor

# Try built-in commands
opencli list
opencli hackernews top --limit 5
opencli bilibili hot --limit 5

# Install AI agent skills
npx skills add jackwener/opencli

# Tell your agent: "Help me check my Xiaohongshu notifications"

For Whom Is OpenCLI?

Perfect for you if:

• You work with Chinese platforms (Bilibili, Zhihu, Xiaohongshu) and want CLI access
• You run AI agents and want them to interact with websites without expensive vision tokens
• You want to reuse your Chrome login session across automation workflows
• You need to operate Electron apps (Cursor, Notion, etc.) from the terminal
• You want zero-cost deterministic adapters that don’t require browser at runtime

Browser-use is better if:

• You want a fully autonomous AI agent that plans and executes multi-step tasks
• You need screenshot-based visual understanding
• You’re building a product rather than a personal workflow tool
• You want the most battle-tested solution (89K stars vs 16K)

The Bigger Picture: CLI as the Agent Interface

OpenCLI is part of a larger trend: CLI becoming the de facto interface for AI agents. Just as developers migrated from GUI to CLI for efficiency, AI agents are now doing the same — and OpenCLI makes that transition seamless for both humans and agents operating the same websites.

The fact that OpenCLI reuses Chrome’s logged-in state is especially important for agents: it means the agent operates with your permissions, your cookies, your 2FA sessions. No need to build auth integrations. No need to store credentials. Just open the site and go.

As the OpenCLI README puts it: “Turn websites, browser sessions, Electron apps, and local tools into deterministic interfaces for humans and AI agents.”

OpenCLI is available at github.com/jackwener/OpenCLI and npm @jackwener/opencli. Docs at opencli.info. Apache 2.0 licensed.