Turn any website into a production-grade CLI — automatically.
Quick Start •
Examples •
How It Works •
Protocols •
Registry •
Roadmap •
Contributing
---
**CLI-Anything-Web** is a [Claude Code](https://docs.anthropic.com/en/docs/claude-code) plugin that generates production-grade Python CLIs for **any** web application by capturing its live HTTP traffic. Point it at a URL, and get a fully working CLI on your PATH — with auth, REPL mode, `--json` output, and tests.
> [!WARNING]
> **Experimental Project — Use Responsibly**
>
> This project uses **undocumented web APIs** reverse-engineered from live HTTP traffic. These APIs can change without notice.
>
> - **Not affiliated with any website** — This is an independent open-source project
> - **APIs may break** — Websites can change their internal endpoints, HTML structure, or add protections at any time
> - **Respect rate limits** — Generated CLIs include exponential backoff, but heavy usage may be throttled
> - **For personal use** — Best suited for prototyping, automation, research, and personal productivity
>
> Generated CLIs interact with real production services. Use them responsibly and in accordance with each website's terms of service.
## 💡 The Idea
Most web apps don't have public APIs. **CLI-Anything-Web** changes that:
> **No API docs needed. No reverse-engineering by hand. Just point and generate.**
## 🚀 Quick Start
### Prerequisites
| Requirement | Version | Why |
|------------|---------|-----|
| [Claude Code](https://docs.anthropic.com/en/docs/claude-code) | With plugin support | Runs the generation pipeline |
| [Node.js](https://nodejs.org/) | 18+ | For playwright traffic capture |
| [Python](https://python.org/) | 3.10+ | Generated CLIs are Python |
### Install the Plugin
```bash
# Inside Claude Code
/plugin marketplace add ItamarZand88/CLI-Anything-WEB
/plugin install cli-anything-web
/reload-plugins
```
### Generate Your First CLI
```bash
/cli-anything-web https://your-favorite-website.com
```
The agent opens a browser, asks you to log in if needed, captures traffic, and generates a complete CLI. That's it.
## 📦 Examples
20 real CLIs generated by the plugin — shipped as reference implementations:
| CLI | Website | Protocol | Auth | Skill | Description |
|-----|---------|----------|------|-------|-------------|
| [`cli-web-futbin`](futbin/) | FUTBIN | HTML + JSON API | None | [📖 Skill](.claude/skills/futbin-cli/SKILL.md) | EA FC player database — search, compare, prices, market analysis, arbitrage, trading signals |
| [`cli-web-notebooklm`](notebooklm/) | Google NotebookLM | batchexecute RPC | Google SSO | [📖 Skill](.claude/skills/notebooklm-cli/SKILL.md) | Notebooks, sources, chat, 9 artifact types (audio, video, slides, quiz, mindmap) |
| [`cli-web-gh-trending`](gh-trending/) | GitHub Trending | HTML scraping | None | [📖 Skill](.claude/skills/gh-trending-cli/SKILL.md) | Trending repos & developers with language/time filters |
| [`cli-web-producthunt`](producthunt/) | Product Hunt | Next.js RSC flight (curl_cffi) | None | [📖 Skill](.claude/skills/producthunt-cli/SKILL.md) | Today's launches, leaderboards, product details |
| [`cli-web-unsplash`](unsplash/) | Unsplash | REST JSON (camoufox browser) | None | [📖 Skill](.claude/skills/unsplash-cli/SKILL.md) | Photo search, download, topics, collections, profiles |
| [`cli-web-booking`](booking/) | Booking.com | GraphQL + HTML (curl_cffi) | WAF cookies | [📖 Skill](.claude/skills/booking-cli/SKILL.md) | Hotel search, property details, destination resolution |
| [`cli-web-stitch`](stitch/) | Google Stitch | batchexecute RPC | Google SSO | [📖 Skill](.claude/skills/stitch-cli/SKILL.md) | AI UI design — generate mobile/web app mockups from text prompts |
| [`cli-web-pexels`](pexels/) | Pexels | SSR + __NEXT_DATA__ (curl_cffi) | None | [📖 Skill](.claude/skills/pexels-cli/SKILL.md) | Free stock photos & videos — search, download, collections, profiles |
| [`cli-web-reddit`](reddit/) | Reddit | REST JSON API (curl_cffi) | Optional (OAuth) | [📖 Skill](.claude/skills/reddit-cli/SKILL.md) | Feeds, subreddits, search, vote, comment, submit, save, inbox |
| [`cli-web-gai`](gai/) | Google AI Mode | Browser-rendered (Playwright) | None | [📖 Skill](.claude/skills/gai-cli/SKILL.md) | AI-powered search with source references |
| [`cli-web-youtube`](youtube/) | YouTube | InnerTube REST API (httpx) | None | [📖 Skill](.claude/skills/youtube-cli/SKILL.md) | Search videos, video details, trending, channel info |
| [`cli-web-hackernews`](hackernews/) | Hacker News | REST API — Firebase + Algolia (httpx) | Cookie (optional) | [📖 Skill](.claude/skills/hackernews-cli/SKILL.md) | Stories, search, comments, users, upvote, submit, comment, favorite |
| [`cli-web-codewiki`](codewiki/) | Google Code Wiki | batchexecute RPC | None | [📖 Skill](.claude/skills/codewiki-cli/SKILL.md) | AI-generated repo docs, wiki sections, Gemini chat, download as .md |
| [`cli-web-chatgpt`](chatgpt/) | ChatGPT | REST API + Camoufox browser | Browser (OpenAI SSO) | [📖 Skill](.claude/skills/chatgpt-cli/SKILL.md) | Ask questions, generate/download images, conversations, models |
| [`cli-web-airbnb`](airbnb/) | Airbnb | SSR HTML + niobeClientData (curl_cffi) | None | [📖 Skill](.claude/skills/airbnb-cli/SKILL.md) | Search stays, listing details, amenities, host info, autocomplete locations |
| [`cli-web-amazon`](amazon/) | amazon.com | SSR HTML + REST JSON (curl_cffi) | None | [📖 Skill](.claude/skills/amazon-cli/SKILL.md) | Search products, view details, browse bestsellers |
| [`cli-web-tripadvisor`](tripadvisor/) | TripAdvisor | SSR HTML + JSON-LD (curl_cffi) | None | [📖 Skill](.claude/skills/tripadvisor-cli/SKILL.md) | Search locations, hotels, restaurants, and attractions |
| [`cli-web-linkedin`](linkedin/) | LinkedIn | GraphQL + Voyager REST (curl_cffi) | Cookie | [📖 Skill](.claude/skills/linkedin-cli/SKILL.md) | Search, feed, profiles, jobs, posts, reactions, comments, network, messaging (26 cmds) |
| [`cli-web-capitoltrades`](capitoltrades/) | Capitol Trades | SSR HTML + BFF JSON (curl_cffi) | None | [📖 Skill](.claude/skills/capitoltrades-cli/SKILL.md) | US congressional stock trades (STOCK Act) — trades, politicians, issuers, price history |
| [`cli-web-worldcup`](worldcup/) | ESPN / FIFA World Cup 2026 | REST JSON (ESPN + The Odds API) | None | [📖 Skill](.claude/skills/worldcup-cli/SKILL.md) | FIFA World Cup 2026 — fixtures, nations, squads, group standings, and bookmaker odds (read-only) |
### Try them yourself
```bash
# Install the entire fleet (every CLI) in one command:
pip install cli-anything-web # or: uv tool install cli-anything-web
# ...or install only the CLIs you want, by name (see the table above):
pip install cli-web-futbin
uvx cli-web-futbin --help # or run it once, no install, via uvx
```
> Working from a clone instead? Install a CLI editable with `pip install -e /agent-harness` (see [QUICKSTART](QUICKSTART.md)).
```bash
# Required browsers
playwright install chromium # for CLIs that require browser-based auth or rendering
python -m camoufox fetch # for CLIs that use stealth browser (anti-bot bypass)
```
**GitHub Trending** — no auth, great first test
```bash
pip install cli-web-gh-trending
cli-web-gh-trending repos list --language python --since weekly --json
```
**FUTBIN** — search EA FC players
```bash
pip install cli-web-futbin
cli-web-futbin players search --name "Messi" --json
```
**NotebookLM** — requires Google login
```bash
pip install cli-web-notebooklm
cli-web-notebooklm auth login
cli-web-notebooklm notebooks list --json
```
**Product Hunt** — no auth, bypasses Cloudflare
```bash
pip install cli-web-producthunt
cli-web-producthunt posts list --json
```
**Unsplash** — photo search
```bash
pip install cli-web-unsplash
cli-web-unsplash photos search "mountains" --json
```
**Booking.com** — hotel search
```bash
pip install cli-web-booking
cli-web-booking auth login # required — fetches WAF browser cookies
cli-web-booking search find "Paris" --json
```
**Google Stitch** — requires Google SSO login
```bash
pip install cli-web-stitch
cli-web-stitch auth login
cli-web-stitch projects list --json
```
**Pexels** — free stock photos & videos
```bash
pip install cli-web-pexels
cli-web-pexels photos search "mountains" --json
```
**Reddit** — feeds, search, vote, comment, submit
```bash
pip install cli-web-reddit
cli-web-reddit auth login # optional — required for voting, commenting, submitting
cli-web-reddit feed hot --limit 5 --json
```
**Hacker News** — stories, search, upvote, submit
```bash
pip install cli-web-hackernews
cli-web-hackernews auth login # optional — required for upvoting, submitting, commenting
cli-web-hackernews stories top -n 5 --json
```
**Google Code Wiki** — AI-generated repo docs + Gemini chat
```bash
pip install cli-web-codewiki
cli-web-codewiki wiki sections excalidraw/excalidraw --json
cli-web-codewiki chat ask "How does rendering work?" --repo excalidraw/excalidraw --json
```
**Airbnb** — search stays, get listing details
```bash
pip install cli-web-airbnb
cli-web-airbnb search stays "London, UK" --checkin 2024-06-01 --checkout 2024-06-05 --adults 2 --json
cli-web-airbnb listings get 770993223449115417 --json
cli-web-airbnb autocomplete locations "New Yor" --json
```
**TripAdvisor** — search hotels, restaurants, and attractions (no auth, DataDome bypass)
```bash
pip install cli-web-tripadvisor
cli-web-tripadvisor hotels search "Paris" --json
cli-web-tripadvisor attractions search "London" --json
```
**ChatGPT** — ask questions, generate images
```bash
pip install cli-web-chatgpt
cli-web-chatgpt auth login # optional — required for image generation & conversations
cli-web-chatgpt chat ask "Explain quantum computing in one sentence" --json
cli-web-chatgpt chat image "A sunset over mountains" --output sunset.png --json
```
**Google AI Mode** — AI-powered search (requires Playwright browser)
```bash
pip install cli-web-gai
playwright install chromium
cli-web-gai search ask "What is quantum computing" --json
```
**Amazon** — search products, bestsellers, autocomplete (no auth required)
```bash
pip install cli-web-amazon
# Search Amazon products
cli-web-amazon search "laptop" --json
# View product details
cli-web-amazon product get B0GRZ78683 --json
# Browse Best Sellers
cli-web-amazon bestsellers electronics --json
```
**LinkedIn** — search people/jobs/companies, feed, profiles, post, network, messaging (auth required)
```bash
pip install cli-web-linkedin
cli-web-linkedin auth login # required — browser-based LinkedIn SSO
# Search people and jobs
cli-web-linkedin search people "python developer" --limit 5 --json
cli-web-linkedin jobs search "software engineer" --json
# View feed, profile, connections
cli-web-linkedin feed --count 5 --json
cli-web-linkedin profile me --json
cli-web-linkedin network connections --limit 10 --json
```
**Capitol Trades** — US congressional stock trades (no auth, CloudFront bypass)
```bash
pip install cli-web-capitoltrades
# Latest trades across Congress
cli-web-capitoltrades --json trades list --page-size 20
# Trades for a specific ticker
cli-web-capitoltrades --json trades by-ticker NVDA
# Top politicians by trade volume
cli-web-capitoltrades --json politicians top --by trades --page-size 10
# Rich issuer data with 1-year price history
cli-web-capitoltrades --json issuers search nvidia
```
### Agent-Native: Claude uses the CLIs automatically
Every generated CLI comes with a **skill file** (`.claude/skills/-cli/SKILL.md`) that Claude Code loads automatically. The skill tells Claude exactly how to use the CLI — every command, every option, every output format. When you ask Claude a question like *"find me a hotel in Paris"*, it reads the booking skill and runs `cli-web-booking search find "Paris"` on its own.
Each skill includes:
- All available commands with arguments and options
- Output format for each command
- Auth requirements and setup instructions
- Common agent patterns (piping, filtering, chaining commands)
### MCP Server Mode: every CLI is also an MCP server
Every generated CLI doubles as a **Model Context Protocol server** — run
`cli-web- mcp-serve` and any MCP client (Claude Desktop, IDEs, other
agents) gets every command as a typed MCP tool, with input schemas derived
from the CLI's own options and the standard `--json` envelope as tool output:
```jsonc
// e.g. in an MCP client config
{
"mcpServers": {
"hackernews": { "command": "cli-web-hackernews", "args": ["mcp-serve"] }
}
}
```
No extra dependencies, no separate server code — the adapter walks the Click
command tree at startup, so new commands automatically become new tools.
Every CLI also ships a `doctor` command (`cli-web- doctor [--json]`)
that diagnoses the local setup — install, Python version, auth file and its
permissions, optional dependencies — before you ever need to open an issue.
## ⚙️ How It Works
The plugin runs a 4-phase pipeline, fully automated by Claude:
| Phase | What Happens |
|-------|-------------|
| **1. Capture** | Opens browser via Playwright, records all HTTP traffic while you use the site |
| **2. Analyze & Generate** | Identifies protocol (REST/GraphQL/RPC/HTML), maps endpoints, generates full Python CLI |
| **3. Test** | Writes unit tests (mocked) + E2E tests (live API) + subprocess tests |
| **4. Publish** | `pip install -e .` → on your PATH → smoke tested → Claude skill generated |
## 🔧 What Every Generated CLI Includes
| Feature | Details |
|---------|---------|
| **Click commands** | `cli-web- [options]` |
| **Interactive REPL** | Run with no args — history, autocomplete, branded prompt |
| **`--json` output** | Machine-readable on every command — pipe into `jq`, scripts, or agents |
| **Auth management** | Python Playwright browser login → cookie extraction → `auth.json` |
| **Error handling** | Typed exceptions → structured JSON errors: `{"error": true, "code": "AUTH_EXPIRED"}` |
| **Tests** | Unit (mocked HTTP) + E2E (live API) + subprocess (`_resolve_cli`) |
| **Installable** | `pip install -e .` puts it on your PATH immediately |
## 📋 Commands
| Command | Description |
|---------|-------------|
| `/cli-anything-web ` | Full pipeline — capture, analyze, generate, publish |
| `/cli-anything-web:record ` | Capture traffic only (for exploration) |
| `/cli-anything-web:refine ` | Add more commands to an existing CLI |
| `/cli-anything-web:test ` | Run tests and update TEST.md |
| `/cli-anything-web:validate ` | Validate against 75-check quality standards |
| `/cli-anything-web:list` | List all generated CLIs |
## 🌐 Supported Protocols
The plugin auto-detects and handles multiple web architectures:
| Protocol | Example Sites | How It's Handled |
|----------|--------------|-----------------|
| **REST / JSON API** | Monday.com, Dev.to | `httpx` client, JSON response parsing |
| **Server-rendered HTML** | GitHub, FUTBIN, Hacker News | `BeautifulSoup4` + CSS selectors |
| **Anti-bot (curl_cffi)** | Product Hunt, Reddit, Amazon | `curl_cffi` with Chrome TLS impersonation |
| **JS proof-of-work challenge** | Unsplash, ChatGPT | `camoufox` (stealth headless Firefox) solves the challenge |
| **GraphQL** | Shopify, GitHub API v4 | Query abstraction into CLI commands |
| **GraphQL + AWS WAF** | Booking.com | `curl_cffi` + WAF cookie bypass |
| **Google batchexecute** | NotebookLM, Google Docs, Keep | Custom RPC encoder/decoder |
| **Browser-rendered (Playwright)** | Google AI Mode | Headless browser rendering + content extraction |
## 🏗️ Repository Structure
```
CLI-Anything-WEB/
├── cli-anything-web-plugin/ # 🔌 The installable Claude Code plugin
│ ├── commands/ # Slash command definitions
│ ├── skills/ # 4-phase pipeline skills
│ │ ├── capture/ # Phase 1: browser + traffic capture
│ │ ├── methodology/ # Phase 2: analysis + code generation
│ │ ├── testing/ # Phase 3: test generation
│ │ └── standards/ # Phase 4: validation + publishing
│ ├── scripts/ # Shared utilities
│ └── HARNESS.md # Complete methodology SOP
│
├── stitch/ # 🎨 Google Stitch (batchexecute RPC)
├── reddit/ # 💬 Reddit (REST JSON API)
├── booking/ # 🏨 Booking.com (GraphQL + AWS WAF)
├── gai/ # 🤖 Google AI Mode (Browser-rendered)
├── notebooklm/ # 📓 NotebookLM (Google batchexecute)
├── pexels/ # 📸 Pexels (SSR + __NEXT_DATA__)
├── unsplash/ # 📷 Unsplash (REST JSON + camoufox)
├── producthunt/ # 🚀 Product Hunt (Next.js RSC + curl_cffi)
├── futbin/ # 🎮 FUTBIN (HTML + JSON)
├── gh-trending/ # 📈 GitHub Trending (HTML scraping)
├── youtube/ # 🎬 YouTube (InnerTube REST API)
├── hackernews/ # 📰 Hacker News (Firebase + Algolia API)
├── codewiki/ # 📚 Google Code Wiki (batchexecute RPC)
├── chatgpt/ # 🤖 ChatGPT (REST API + Camoufox)
├── airbnb/ # 🏠 Airbnb (SSR HTML + niobeClientData)
├── amazon/ # 🛒 Amazon (SSR HTML + REST JSON)
├── tripadvisor/ # 🌍 TripAdvisor (SSR HTML + JSON-LD, curl_cffi)
├── linkedin/ # 💼 LinkedIn (GraphQL + Voyager REST, curl_cffi)
└── capitoltrades/ # 🏛️ Capitol Trades (SSR HTML + BFF JSON, curl_cffi)
```
## 🔗 Inspiration
This project is directly inspired by [**CLI-Anything**](https://github.com/HKUDS/CLI-Anything) by HKUDS — a Claude Code plugin that makes **desktop software** (GIMP, Blender, LibreOffice, OBS Studio) agent-native by analyzing source code and generating CLI wrappers.
**CLI-Anything-Web** extends the same vision to the **web** — where there's no source code to analyze, only live HTTP traffic to capture.
| | [**CLI-Anything**](https://github.com/HKUDS/CLI-Anything) | **CLI-Anything-Web** |
|---|---|---|
| **Target** | Desktop apps (GIMP, Blender, OBS) | Web apps (NotebookLM, Booking.com, any website) |
| **Input** | Source code, GUI APIs, plugin systems | Live HTTP traffic from browser |
| **Analysis** | Static code analysis + API mapping | Network traffic capture + protocol detection |
| **Auth** | N/A (local software) | Browser login, cookies, WAF bypass, API keys |
| **Output** | Click CLI + REPL + `--json` + tests | Click CLI + REPL + `--json` + tests |
Together they cover the full spectrum: **CLI-Anything** for desktop, **CLI-Anything-Web** for the web.
## 🗺️ What's Next
We're actively building more CLIs and improving the plugin:
- 🎯 **More CLIs** — Jira, Notion, Monday.com, Spotify
- 🎵 **Content generation** — Suno, ElevenLabs, Midjourney
- 🧠 **Smarter analysis** — Auto-detect auth flows, pagination, WebSocket streams
- 🔄 **CI/CD integration** — Run CLIs in GitHub Actions with env-var auth
- 📦 **Community registry** — Share and install CLIs built by other users
**Want a specific website?** [Open an issue](https://github.com/ItamarZand88/CLI-Anything-WEB/issues/new) with the URL — we'll prioritize the most requested ones.
## 🤝 Contributing
We'd love your help! See **[CONTRIBUTING.md](CONTRIBUTING.md)** for the full guide.
The quickest ways to contribute:
| What | How |
|------|-----|
| **Build a new CLI** | Run `/cli-anything-web ` on any website, submit a PR |
| **Improve the plugin** | Add patterns to `cli-anything-web-plugin/skills/` |
| **Report bugs** | [Open an issue](https://github.com/ItamarZand88/CLI-Anything-WEB/issues) with `--json` output |
| **Fix a broken CLI** | Website changed? Update the scraper/client |
## 📄 License
This project is licensed under the [MIT License](LICENSE).
---
Built with Claude Code ·
Inspired by CLI-Anything ·
Request a CLI