Most “agents” today are just wrappers around a loop of prompts. I wanted something more ambitious: an actual cognitive engine.

So I built AGI‑SaaS: a modular AGI system you can use like a brain, or deploy like an API.

Here’s what makes it different:

Plugin-based cognition Add memory, reasoning, search, tools — as Python plugins.

LLM-agnostic OpenAI, Claude, LLaMA, OpenRouter — switch in seconds.

Mental journal Trace every thought and decision. Debug cognition itself.

FastAPI-ready Turn it into an agent, a SaaS backend, or your own AGI product.

AGI isn’t about bigger models. It’s about better architecture.

Very nice project, and I agree that the path to AGI is about architecture, not just model size.

But this isn’t AGI. I don’t think we should dilute the AGI term to advertise projects/products.

I feel like it’s painfully obvious that we’re lacking something fundamental in the neural net architectures we use, that makes it so there’s aspects of general intelligence that even a bird possesses, that our AIs don’t.