DEV Community

Cover image for I Added MCP Integration So ChatGPT Can Deploy My Code (While I Walk Away From Docker & YAML) 💥
Harun - solo dev
Harun - solo dev

Posted on

I Added MCP Integration So ChatGPT Can Deploy My Code (While I Walk Away From Docker & YAML) 💥

#showdev #ai #mcp #buildinpublic

Two days ago, I published an article about how I tried to hack my own PaaS and fixed the critical security bugs I found.

Today, I’m back to tell you that ChatGPT can now deploy my code for me.

Welcome to HYNA-Stage v3.0.0.


(Yes, this is literally how I feel every time a sandbox provisions in 7 seconds without a single docker-compose.yml file in sight).

🤖 THE "AHA!" MOMENT: WHAT IS MCP?

If you haven't heard of MCP (Model Context Protocol), here is the simple version: it’s the new standard that lets AI assistants like ChatGPT and Claude Desktop actually use tools instead of just talking about them.

Before v3.0.0:
I ask ChatGPT: "How do I deploy a Node.js app?"
It gives me a 10-step tutorial and a bunch of YAML.

After v3.0.0:
I ask ChatGPT: "Deploy my Node.js app."
It actually deploys it. It talks directly to HYNA-Stage’s MCP server, provisions an isolated E2B MicroVM, and gives me the live URL.

No copy-pasting. No terminal commands. No YAML explosions. Just pure, AI-native infrastructure.

🚧 THE INDIE-HACKER FLEX: BYPASSING THE PAYWALL

Here is the frustrating reality: OpenAI’s public MCP directory requires a payment method for individual verification.

As a 12-year-old founder building on a ₹0 budget from a $150 Android phone, I don’t have a credit card to verify with. Big Tech gatekeeping at its finest.

So we didn't let that stop us.

You don't need the official directory to use this. Anyone can add HYNA-Stage as a custom connector right now:

  1. Go to ChatGPT Plugins → Create Plugin → With MCP
  2. Enter server URL: https://hyna-stage.indevs.in
  3. Complete our custom OAuth 2.1 + PKCE flow
  4. Done. Full access. No paywall. No gatekeeping.

Custom connectors are how the real builders operate anyway.

⚡ THE NUMBERS THAT ACTUALLY MATTER

I won't bore you with a massive architecture diagram (you can find the full v3.0.0 technical report on our GitHub). But here is what you need to know about the performance:

  • Warm Sandbox Claims: Under 3 seconds. (Cold starts are 7-15s).
  • Hot Reload Latency: ~300ms. Change a line of code, see it live instantly.
  • Uptime: 99.9% over the last 7 days.
  • Deployment Success Rate: >99% (thanks to the security hardening we did two days ago).

And yes, every single line of this MCP server, OAuth flow, and Cloudflare proxy was written and deployed from Termux on my phone.

🛠️ HOW TO TRY IT RIGHT NOW

1. The AI Way (ChatGPT / Claude Desktop)

  • Go to chatgpt.com/plugins
  • Select Create Plugin → With MCP
  • Server URL: https://hyna-stage.indevs.in
  • Authenticate and start telling the AI to deploy your code.

2. The Classic Way (CLI)

npm install -g @hynaweb/stage
hyna-stage login
hyna-stage init node
Enter fullscreen mode Exit fullscreen mode

🧠 THE REAL TAKEAWAY

The future of development isn't wrestling with Docker networks or debugging YAML indentation at 2 AM.

The future is AI-native, instant, isolated, and accessible to anyone with a phone.

To everyone who read the security audit, to Akhouri (my CTO) for debugging the OAuth flow with me, and to this community for treating me like a peer: This v3.0.0 is for you.

We are just getting started.

Try it out, break it, and let me know what you build! 🐯☁️

— Harun, Founder & CEO, HYNAWEB

Top comments (0)