HN Showcase
Back to browse
Build and Host AI apps on your own servers

Build and Host AI apps on your own servers

by winash83·Jun 22, 2026·4 points·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemSlick

Self-hosted runtime lets AI agents deploy apps without opening inbound ports.

Strengths
  • •MCP integration allows any compliant agent to drive the deployment pipeline.
  • •Zero inbound ports required reduces attack surface for internal tooling significantly.
  • •Keeps data on user hardware, solving privacy concerns with cloud AI builders.
Weaknesses
  • •Relies on AI agents writing production-ready code, which remains hit-or-miss today.
  • •Requires users to manage their own server infrastructure and Docker updates.
Target Audience

Developers building internal tools with AI agents

Similar To

E2B · Replit · Lovable

Post Description

I wanted a way to build and host some vibecoded apps/dashboards etc , but figured there was no good way outside using things like lovable and replit , where the code and data and the hosting lives on their servers

Many of the online app builders don't allow you to switch the model to an open model and charge premiums for tokens, Some of the app builders even want you to paste DB credentials in the chat which to me was a complete no no

I have been working on an alternative to this model , where you bring your own servers and use any AI harness to build and host apps

You bring 3 things

1. A server - even your laptop will do or a $5 VPS - should run docker 2. AI harness - anything - Cursor , roo , Claude , Pi - It should speak MCP 3. AI Api key

You use an mcp in stdio mode to drive a remote sandbox on your own server , and then with one click , just deploy the app publicly

The backend is on docker and I plan to extend it to other backends soon

You can build anything that you can build on Lovable or Replit , even full stack apps , You can bring your databases, bind env variables and integrate almost anything without the secrets ever leaving your machine or your server

Everything is TLS end to end, Your server needs no inbound ports

I would love some feedback on this, The core is open source , The hosted service is free to use

Here is something I built today

A real-time operations dashboard for research vessels, combining live tracking, mission planning, and environmental monitoring in a unified interface (https://vessel-ops-dashboard-4c3976e79335c1aa.agentry.live/)

Similar Projects

AI/ML●Mid

Docker AI Stack – Deploy 8 self-hosted AI services with one command

Yet another one-command AI stack when Ollama and Jan already exist.

Ship It
hwdsl2
101mo ago
Infrastructure●●Solid

Deploy OpenClaw on your Own server in one click

A pragmatic product: it strips away the ops friction of running OpenClaw by wiring DigitalOcean OAuth + a launcher to get an instance live in minutes, and even advertises Telegram and multi-LLM support. That said, it’s not reinventing the wheel — the real value is in the UX for non-operators and the security pitch about keeping keys off hosted services; I’d want clearer docs on secret handling, upgrades, and which parts are open source before trusting it with production data.

Ship ItSolve My Problem
pushkar_aditya
214mo ago