Deploy code from ChatGPT via MCP integration
A developer added MCP integration to their self-built PaaS so ChatGPT can deploy their code without Docker or YAML. ·
Done withChatGPT
- What they did
- Harun built an MCP server into his PaaS, HYNA-Stage v3.0.0. ChatGPT or Claude Desktop connects to it as a custom connector, using a custom OAuth 2.1 + PKCE flow, because OpenAI's public directory needs a payment method. After that, asking ChatGPT to deploy an app provisions an isolated E2B MicroVM and returns a live URL. The server runs on Cloudflare and Supabase, and he wrote it from Termux on his phone.
- How it went
- He reports warm sandbox claims under 3 seconds, cold starts of 7-15 seconds, about 300ms hot reload, 99.9% uptime over 7 days, and over 99% deployment success. These figures are self-reported, with no methodology given.
- Worth knowing
- The free tier caps users at 3 concurrent sandboxes, and the stack runs on free tiers of Supabase, Cloudflare, E2B and Netlify, which he says he will outgrow.
Try it yourself with ChatGPT
Help me set up deployment from my AI assistant for my project in [folder or repo]. Create an MCP server that exposes a deploy tool, which builds and ships the app to [hosting target] without needing me to write Docker or YAML by hand. Show me the exact commands and config before running any deploy, and stop when I can trigger a test deploy from my assistant.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: DEV · Nov 10, 2026
Related use cases
Generate LEGO CAD models in LDraw with AI agents
A toolset and docs let ChatGPT and Claude agents write LDraw source files that become LEGO CAD models, packaged as a dockerized web app.
Draft parent-teacher communications
A teacher says ChatGPT handled all their parent-teacher communications for a month, and they also tried it for lesson planning.
Automate technical blog post production
The Basilica team built a Make and ChatGPT workflow to produce technical blog posts faster, reporting about 6 hours saved per post.
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Monitor tasks and open pull requests with long-running agents
A commenter runs a persistent script that watches tasks and uses Cursor to open pull requests, building a documentation summarizer and a data pipeline.
Write a tricky Flutter function with Copilot
The poster had Copilot write a difficult Flutter function for their app, and a commenter turns written requirements and code samples into working output, then builds their own test harness.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.