Self-hosted multi-agent AI workspace — looking for honest feedbackSelf-hosted multi-agent AI workspace — looking for honest feedback

Hey MongoDB :waving_hand:

I’ve been building AladdinAI — a self-hosted, multi-agent AI workspace.

Open source, runs entirely in your own infrastructure, bring-your-own LLM provider.

What’s inside:

  • Multi-agent system with persistent memory (private per-agent + shared)

  • Vector recall via MongoDB Atlas

  • Provider-agnostic LLM layer — connect NIM, OpenAI, Anthropic, or local models

  • Tools registry with @tool decorator (no LangChain)

  • Safety stack: NemoGuard, GLiNER PII, Llama Guard

  • Channels: Telegram, WhatsApp, Email, SMS

  • Terminal-in-browser (ttyd + wetty) with Traefik routing

  • Triggers/automations via APScheduler

Stack: FastAPI + Next.js 15 + Postgres + MongoDB.

Setup is one command:

npx aladdin-ai

After that you add your LLM provider through the UI (NIM has a free tier

that works well, but anything OpenAI-compatible will plug in).

The honest part:

436 cloners organically, zero feedback. I genuinely don’t know if people

are loving it, hating it, or just curious about the README. That’s why

I’m here — I’d rather hear “this is broken” than silence.

If you try it, even just \npx aladdin-ai\ and click around for 5 minutes,

please tell me:

  • What worked?

  • What broke?

  • What did you expect that wasn’t there?

Brutal honesty welcome.

— Aladdin