OmniMind AI

• 

  OmniMind AI logo

Inspiration

I switch between AI models constantly when I build. ChatGPT for one thing, Claude for another, Gemini when I want a second opinion. The problem is each model only has memory of its own conversations. ChatGPT knows nothing about what I told Claude last session. Claude knows nothing about the decisions I made in ChatGPT. Every switch means starting from zero again.

I wanted a way to carry my project context across all of them, so switching models does not mean losing everything the other one knew.

What it does

OmniMind AI is a dashboard that acts as shared memory across all your AI models. You define your project once inside OmniMind, your stack, your goals, your constraints, what you are working on right now. Then when you switch from ChatGPT to Claude, or from Claude to Grok, OmniMind generates a context block in the exact format that model expects so it picks up exactly where the last one left off.

It also includes a Context Debt Score, a live 0 to 100 gauge that shows how much of your project the current model is missing. A Token Estimator that shows the API cost of your context per model before you export. And a Chrome extension concept that captures context from your AI conversations passively in the background, so when you switch models it already has everything ready to transfer.

How we built it

Built solo over 3 days using Next.js 15, TypeScript, Tailwind CSS, and shadcn/ui. The core logic is a pure function context engine that handles scoring, token estimation, and export generation separately from the UI. The Chrome extension was designed as a three state prototype showing active capture, transfer ready, and idle.

Challenges we ran into

The hardest part was making the problem feel real in the UI, not just describing it. Context drift is invisible, you do not notice it until you have re-explained your project for the tenth time that week. The Context Debt Score was the solution to that. Giving the problem a number made it tangible.

Accomplishments that we're proud of

Getting the export engine to generate meaningfully different output per model. Claude gets a structured CLAUDE.md file. Cursor gets .cursorrules format. ChatGPT gets a system prompt paragraph. Same project, three different formats, one click.

And the Context Debt Score. I have not seen that framing anywhere else and it feels like the kind of thing that would actually change how a developer thinks about their AI workflow.

What we learned

The real product insight was not about saving time. It was about the gap between what your AI model knows and what your project actually is. Once I understood that as the core problem, every feature became obvious.

What's next for OmniMind AI

A fully working Chrome extension that reads your conversations across ChatGPT, Claude, and Gemini, extracts context automatically, and syncs it to your dashboard. So when you switch models mid-project, the new one already has everything the last one knew. No copy pasting, no re-explaining. Just switch and keep building.

Built With

• next.js-15
• recharts
• shadcn/ui
• tailwind-css
• typescript
• vercel