DEV Community: MR-RAHAD

I Wrapped All My APIs Behind One Key and Started Selling Access (From Bangladesh)

MR-RAHAD — Fri, 09 Oct 2026 15:44:53 +0000

I maintain a handful of small APIs: a Grok chat bot, a Kimi chat API, an image generator, a photo enhancer, and a video downloader. Each had its own URL, its own auth (or none), and its own way of breaking at 2 AM.

So I built a gateway. One endpoint, one API key, plan limits, and a dashboard. It's live at Rahad API Marketplace (Railway), and this is what I learned shipping it in about a day.

The shape of it

One key (rk-...) for every backend: chat, image generation, photo enhancement, media download
Plan limits that actually stop abuse: Free 100/day, Pro 5000/day, Unlimited 100k/day
Usage is charged only when an upstream call succeeds — if my backend fails you, it shouldn't cost your quota
A dashboard where you sign up, verify your email, and get a free key instantly

Payments are the Bangladesh part of the story: no Stripe here, so Pro upgrades run through bKash with a manual transaction-ID approval flow. Ugly? A little. Working? Completely.

One bug class that bit me

When an upstream API fails (say it returns 401), my first version passed that straight through to the buyer. Their client then told them "unauthorized" — which sounds like their key is wrong. That's my bug wearing their error message.

Fix: the gateway now translates any upstream failure into an honest 502 that names the backend, and only counts usage on success. Small change, much fairer billing.

What it's honest about

GET /v1/models doesn't pretend everything is fine. It probes the upstreams and reports real availability — if a backend is down, the endpoint says so instead of selling you a broken promise.

Try it

Live: https://rahad-api-marketplace-production.up.railway.app
Free key: sign up, verify email, done — 100 requests/day, no card

Honest question for the comments

For those of you selling small APIs from countries without Stripe — how do you take money? Manual mobile-money approvals like mine, Paddle/Lemon Squeezy, or something else entirely? I'm genuinely deciding what to do next, and I'd rather learn from your scars than earn my own.

I Built a ChatGPT API Without Login — How the Anonymous Flow Actually Works

MR-RAHAD — Thu, 08 Oct 2026 21:08:11 +0000

I was curious how ChatGPT's logged-out "guest" mode really works under the hood, so I built a small educational project around it: a working API that chats with ChatGPT without any account, login, or API key.

Repository: https://github.com/MR-RAHAD/chatgpt-api

What it does

POST /chat with a prompt returns ChatGPT's reply as JSON. Every request runs a completely fresh anonymous session — the same steps a browser would take, but without a browser.

How it works

Studying the guest flow turned into a tour of modern anti-bot engineering:

Session — the flow is Cloudflare-gated, so the session opens with a Safari TLS fingerprint.
Prepare — the server issues a "Sentinel" challenge: a proof-of-work plus two obfuscated VM bytecode programs.
Proof-of-work — solved in pure Python (an FNV-1a hash puzzle over a base64 config payload).
VM tokens — the bytecode programs (session-observer + turnstile) are XOR-encrypted; I run them in Node/jsdom against the real page environment to produce the required tokens.
Finalize + conversation — the solved challenge is exchanged for a conversation token, the message is sent, and the streamed (SSE) reply is parsed.

Extras

Proxy rotation — a round-robin proxy pool with automatic cooldown for dead proxies (the guest flow is rate-limited per IP, so rotation spreads the load).
Optional API keys — lock your own deployment down with CHATGPT_API_KEYS.

Honest limitations

~20–40 seconds per reply; requests run one at a time.
Guest quota is roughly 10 messages per 5 hours per identity/IP — proxies help, but nothing here is "unlimited".
It wraps a web frontend, so bundle changes can break the solver until it's re-derived.

Why build this?

🎓 Education and research. This project exists to demonstrate how modern web platforms defend themselves — TLS fingerprinting, proof-of-work challenges, and VM-based token systems — and how those mechanisms work in practice. It's unofficial, not affiliated with OpenAI, and the README carries a full disclaimer.

If you learned something from it, a ⭐ on the repo helps others find it. Issues and questions are welcome!

A question for you

If you were building on top of a no-login API like this, what would you use it for — a Telegram bot, a CLI tool, or something else? And what's the first thing you'd worry about breaking?

Drop your take in the comments — I read every one.

One npm Package to Download Videos from TikTok, Instagram, Facebook, YouTube and 7 More

MR-RAHAD — Fri, 02 Oct 2026 09:15:41 +0000

Every project that touches social media eventually needs the same thing: "give me the direct video URL for this link."

I got tired of wiring up a different scraper for every platform — one library for TikTok, another for Instagram, a third that breaks every time YouTube changes something. So I built rahad-all-downloader-v2: one package, 11 platforms, zero setup.

What it does

11 platforms, one API:

TikTok — no-watermark HD video + MP3 audio extraction
Instagram — Reels, posts, IGTV, no login needed
Facebook — HD & SD quality options
YouTube — with automatic fallback for reliability
X (Twitter), Snapchat (Spotlight & Stories), Threads, Pinterest, Likee, CapCut, Kwai

Zero setup: no cookies, no API keys, no login. Install and go.

Built for real apps: batch downloads with concurrency control, auto-retry with backoff, quality preference, progress callbacks.

Modern: TypeScript definitions, ESM + CJS, ~21KB, MIT licensed.

Quick start

npm install rahad-all-downloader-v2

const { alldl } = require('rahad-all-downloader-v2');

// Single download — returns a direct video URL
const v = await alldl('https://www.tiktok.com/@user/video/123456789');
console.log(v.data.videoUrl); // direct .mp4 link, no watermark
console.log(v.data.title);

// Batch with concurrency control (never throws — failed items are reported)
const results = await alldl.batch(
  ['https://www.instagram.com/reel/ABC123/', 'https://youtu.be/xyz'],
  {
    concurrency: 3,
    onItem: (r) => console.log('done:', r.data?.title),
  }
);

// Retry + quality preference
const hd = await alldl(url, { retries: 3, quality: '720p' });

That's it. One function, every platform.

Why I built it

Most downloader libraries solve one platform well and ignore the rest. Real projects — bots, archivers, content tools, automation — need every platform behind a single stable interface. This package is dependency-free, handles failures gracefully (batch mode never throws), and stays updated as platforms change their frontends.

Try it

If it saves you time, a star on GitHub means a lot. Feedback and issues welcome!

Your turn

Which platform do you download from the most — TikTok, Instagram, YouTube, Facebook…? And is there a platform you wish this package supported but doesn't yet?

Tell me in the comments — the most-requested platform goes to the top of my roadmap.