I built a self-hosted MCP server registry and proxy in TypeScript

Diya — Sat, 13 Jun 2026 16:01:36 +0000

The official MCP registry is written in Go and cloud-hosted.
I wanted something I could self-host, written in TypeScript,
that my team could run inside our own infra.

So I built mcp-hub.

What it does

Register any MCP server (SSE, HTTP, stdio transports)
Proxy all tool calls through one endpoint with auth forwarding
Stream every tool call live via WebSocket (Socket.io)
Monitor server health async via BullMQ jobs

Stack

NestJS — API framework
PostgreSQL — server registry + tool call logs
Redis — WebSocket pub/sub + BullMQ transport
BullMQ — async health check queue
Socket.io — real-time log streaming

How it works

curl -X POST http://localhost:3000/api/v1/servers \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "github-mcp",
    "url": "http://my-mcp-server:8080",
    "transport": "http"
  }'

Then proxy any tool call through mcp-hub:

curl -X POST http://localhost:3000/api/v1/proxy/{serverId}/call \
  -H 'Content-Type: application/json' \
  -d '{ "tool": "create_issue", "input": { "title": "Bug" } }'

Every call is logged to PostgreSQL and streamed live over WebSocket.

Run it in one command

git clone https://github.com/DIYA73/mcp-hub
cd mcp-hub
cp .env.example .env
docker compose up

API at http://localhost:3000/api/v1
WebSocket at ws://localhost:3000/logs

Why I built this

I'm running multiple MCP servers locally and needed a central
place to register them, forward calls, and see what's happening
in real time. Nothing TypeScript-native existed.

GitHub: https://github.com/DIYA73/mcp-hub

Feedback welcome — especially if you're running MCP servers in prod.

I built a self-hosted "Git for AI prompts" — version control, test runner & A/B testing (open source)

Diya — Thu, 11 Jun 2026 23:49:31 +0000

If you're building anything with LLMs, you've probably hit this problem: your prompts are scattered across code files, Notion docs, and Slack threads. There's no version history. You can't easily test a change. And you have no idea what's actually running in production.
I built Prompt Vault to fix this — a self-hosted web app that treats your prompts like code.
What it does
Version control — every edit creates a new version. You can activate any version at any time, just like git checkout.
Auto-detected variables — use {{variable}} syntax in your prompts. The UI automatically detects them and lets you fill values before testing.
Test runner — run any prompt version directly against OpenAI from the UI. See the response, latency, and token usage instantly.
A/B testing — create a test between two prompt versions, run both in parallel, and compare outputs side by side.
Multi-user — JWT auth, each user's data is fully isolated.
Self-hosted — Docker Compose for one-command setup. Your prompts stay on your infrastructure.
Tech stack

Backend: NestJS + TypeScript
Database: PostgreSQL + Prisma
Frontend: Next.js 15 + Tailwind CSS
Auth: JWT + bcrypt
AI: OpenAI API

Quick start
bashgit clone https://github.com/DIYA73/prompt-vault
cd prompt-vault/backend
cp .env.example .env

Add your DATABASE_URL and OPENAI_API_KEY

npm install
npx prisma generate
npx prisma migrate dev --name init
npm run start:dev

New terminal

cd ../frontend
npm install
npm run dev
Open http://localhost:3000 — register, create a project, add your first prompt.
Why I built it
Every team I've seen building LLM features manages prompts the same broken way — hardcoded strings in source code, no history, no way to safely iterate. Prompt Vault is the tool I wished existed.
GitHub
⭐ https://github.com/DIYA73/prompt-vault
Would love feedback — especially from anyone managing prompts at scale.

DEV Community: Diya