About

AI got brilliant. Paying for it got complicated.

Minott Technologies built MPortal because the AI market quietly became a marketplace: a dozen frontier models, each priced differently, each better at different work, with prices that move month to month. Most people pick one subscription and overpay for easy questions; the careful ones juggle four tabs and still guess. MPortal replaces the guessing with arithmetic. Open benchmark data decides which models are good enough for your task, live per-token pricing decides which of those is cheapest, and every answer comes back with a receipt.

We don’t sell AI. We make sure you never buy more of it than the job needs.

Open data, not vibes

Model quality comes from public leaderboards and benchmark indices; prices come from the market, live. The routing math is explainable — every decision shows the models considered and why the winner won.

Receipts, not estimates

After every answer: actual tokens, actual dollars, and the price the premium model would have charged. Savings you can audit beat savings we could advertise.

Your ceiling, respected

The weekly budget you set at signup is a promise we keep — as spending nears it, routing shifts to economical models instead of cutting you off mid-week.

FAQ

Fair questions, straight answers.

Which models does MPortal route between?

The current Claude, GPT, Gemini, and Grok families across premium, mid, and fast tiers — eleven models at launch — with an extended catalog (DeepSeek, Qwen, Kimi) available soon.

Do I need my own AI subscriptions or API keys?

No. One MPortal account reaches every model. You set a weekly budget and MPortal spends it as carefully as the routing math allows.

What happens when I hit my weekly budget?

Nothing abrupt. As spending approaches the line, MPortal quietly downgrades to the most economical models that can still do the job — your budget is a ceiling, not a tripwire.

How do you know which model is “good enough”?

Public benchmark data — arena rankings, intelligence and coding indices — scored per task category. Hard tasks demand near-frontier quality; easy ones don’t pay frontier prices. You can always force best-quality mode for a request.

How do I get access?

MPortal is in early access at app.minott.tech.