Okay folks lets dive into a trending repo in the AI / ai-agent / CLI space that caught my eye and I've seen making the rounds on the hub and socials. No shade or hate to the creator - not really interested in who they are or credentials just gonna poke around and do my thing.
TL:DR
65 scrapers in a trenchcoat borrowing your identity. -> What could go wrong. 😂
Verdict: Pretty damn cool, your days are numbered X API...
| What it is | A CLI that turns websites into terminal commands using YAML pipelines, cookie hijacking, and DOM scraping |
| Stars | 8k+ |
| Make No Mistakes Take | Little buggy but pretty useful |
| Who it's for | If you're building AI agents that need structured web data without paying for 15 different API keys |
OpenCLI
Make any website, Electron App, or Local Tool your CLI.
Zero risk · Reuse Chrome login · AI-powered discovery · Universal CLI Hub
https://github.com/jackwener/opencli
The Stats
8k stars — 391 commands across 65+ services — ~4,800 lines of TypeScript — dependencies are lean (commander, chalk, ws, turndown, js-yaml — no bloat)
How It Actually Works
This thing operates in three modes and they are NOT created equal:
1. Public API mode — the good stuff. Declarative YAML pipelines that chain fetch, filter, and map steps. Here's literally all it takes to build a HackerNews command:
yaml pipeline: - fetch: url: https://hacker-news.firebaseio.com/v0/topstories.json - map: id: ${{ item }} - fetch: url: https://hacker-news.firebaseio.com/v0/item/${{ item.id }}.json - filter: item.title && !item.deleted && !item.dead - limit: ${{ args.limit }}
No code. Just declarative data piping. Pretty elegant.
-
Cookie/Browser mode — for login-walled sites like Reddit and Twitter. Hijacks your existing Chrome session via Chrome DevTools Protocol, uses your cookies to hit authenticated API endpoints, and
returns structured data. You don't need a X API key ($100/mo lol) to pull trending topics. Which I think is the strongest play here. The X api is bonkers bad. -
UI Scraping mode — for Electron apps like Discord and Cursor. Connects directly to the running app and scrapes the DOM. This is where things get rough.
After trying it for 15 min
Public commands — flawless
No auth, no config, just works. Fast. The -f json flag for structured output is exactly what an AI agent would want to consume.
Reddit — solid once you get past setup
Had to open Chrome with --remote-debugging-port=9222, install a browser extension, and set an env var. Took about 3 minutes.
But once running, it pulled full authenticated Reddit data including recursive threaded comment trees. I read an entire LocalLLaMA post about the new Qwen 3.5 models from my terminal. Opens up a browswer and uses your authenticated session.
Discord — pain
Getting Discord connected was a whole thing. Kill the app, relaunch with a debug flag that kept failing, multiple attempts. When it finally worked the data was full of duplicates. And the send command literally types your message into the box and just... leaves it there. Doesn't press enter. Could't figure it out, but doesn't matter.
Under The Hood
Had claude run through the code for me
There's genuinely interesting stuff happening. The raw CDP client is different from Playwright or Puppeteer. It's a straight WebSocket talking to Chrome. There's a 156-line stealth module that patches navigator.webdriver, fakes browser plugins, and scrubs automation fingerprints from stack traces. It's making the browser act like a real user, not a bot. Very smart.
The annoying part is it pops the browser open on every action. It can't pull images. Just returns raw URLs. And about half the 391 commands are for Chinese web platforms (bilibili, zhihu, douyin, weixin, xiaohongshu) -> So little bit of Chinese influence there.
The Verdict
If you pre-set up your workflow and logged into everything in the browser you could do some real damage with this. The public API stuff is fast and clean. The Reddit/Twitter cookie hijacking is
genuinely useful.
Some of the features are cleaner than other. Like some of the electron app stuff is buggy and setup is more painful than I'd like.
Will I use it again? Maybe. The browser cookie stuff to replace reddit and X APIs definitely gives this library a spot on my machine.
Report Card
| Category | Rating |
|---|---|
| Overhype Meter | 🟢 Low — doesn't promise what it can't deliver (mostly) |
| Setup Pain | 🟡 Medium — public stuff is instant, browser stuff takes some fiddling |
| Slop Level | 🟢 Low — real code, lean deps, no wrapper energy |
| Usefulness | 🔴 High — the parts that work actually solve real problems |
| Would I Keep It Installed | 🔴 High — public commands and X/Reddit are solid |