Show HN: Ego – A PaaS API for stateful AI personas | Hacker News Reader

Show HN: Ego – A PaaS API for stateful AI personas

soulstore.ai

1 pointjrhatigan5 comments

Hey HN,

Founder here. I'm excited (and a bit nervous) to share what started as a personal project to solve my own frustration.

Like many of you, I've been building with LLMs and kept hitting the same wall: creating AI agents with any kind of deep, consistent personality is incredibly difficult. Managing massive system prompts, trying to enforce behavioral rules, and handling state just felt like a huge distraction from building the actual application. I just wanted a clean API for this.

So, I built it. Ego is a Persona-as-a-Service platform. The idea is to abstract away the complexity of advanced prompt engineering. You choose a persona from a library, set a few parameters like tone and verbosity in a standard REST call, and get back coherent, in-character responses.

The core of it is a hierarchical model for defining personality traits (the tech for this is now patent-pending). The API itself is built with FastAPI and deployed on AWS App Runner.

The live site is a simple waitlist for now. I'm opening up a private beta for the first 250 developers to help me test it, find the rough edges, and shape the roadmap.

You can check out the landing page again here: https://www.soulstore.ai

I'm here all day to answer any questions about the tech stack, the business model, the patent process, or the journey. Would love to get your feedback and hear what you think.

Say what you will about it. MCP surely has to be your primary interface?

Actually, while MCP is a great direction for standardizing context, the "special sauce" of Ego isn't the context-passing itself, but the proprietary engine that uses that context to dynamically generate prompts from a deep personality model.

For the MVP, I had to build the custom state handler to prove out that core persona logic first. Long-term, I'd love for Ego to accept context from standard protocols like MCP—it would make the engine even more versatile.

Appreciate the sharp question!

System Prompt aside and not that critical today, however you look at it you are stuffing context. MCP is good at that and it works in most AI‘s.

You're definitely right for now, MCP is great for 'stuffing context'.

However, the long-term goal of this framework is to make the persona itself dynamic. It's designed to use that context to subtly update a persona's core personality traits over time, creating a more 'life-like' evolving 'virtual soul'.

The cherry-on-top is framing it all as a 'Persona-as-a-Service' API to make the dev workflow straightforward but powerful. It's not for every application, but for anything that needs real character depth - like in gaming or advanced assistants - the goal is to make it simple for any developer to create those deeper, more meaningful interactions that are typically hard to get right.

Reply on news.ycombinator.com