MPortal by Minott Technologies
One login. Every frontier model.
Words, images, video.
MPortal reads each request and routes it to the model that does the job for the least money — Claude, GPT, Gemini and Grok for text, and dedicated models for image and video generation. Every answer comes back with what it actually cost and an estimate of the electricity it used. Start free, no card.
Free at app.minott.tech — no card, no invite. One AI subscription in place of many.
95%
saved on everyday AI questions
76%
saved on expert coding tasks
23
text, image and video models behind one login
$0
to start — no card, no invite
Savings measured by MPortal’s router against live model pricing, September 2026. Eighteen models on every plan; the extended catalogue takes it to twenty-three on Plus.
One box
Ask for words, a picture, or a video. Same box.
You never pick the model. MPortal reads what you asked for and sends it somewhere that can make it.
Words
Eleven text models
Claude, GPT, Gemini and Grok, from the fast cheap tiers to the frontier — plus DeepSeek, Qwen and Kimi on Plus. Writing, code, analysis, translation, a one-line fact. Attach a PDF, a spreadsheet or a screenshot from Basic upward.
Free plan: text prompts, answered from the model’s own knowledge.
Images
Image generation, from free
Ask for a logo, an illustration, a poster or a photo and the request goes to an image model instead of a chat one — Grok Imagine, Nano Banana or GPT Image, whichever is cheapest for the job. FLUX joins them on Plus.
From about $0.025 an image, so the free plan’s weekly allowance covers a few.
Video
Video generation on Plus
Veo and Seedance, priced by the second and routed the same way as everything else. A clip is a different order of cost from a sentence, so it sits on the plan whose weekly budget can absorb it — and you see the price before it runs.
From about $0.12 for a five-second clip. Plus plan.
How it works
How the AI model router works, in three steps
01 — Read
The task gets understood
MPortal classifies every prompt: coding, writing, math, translation, a quick fact, a picture, a clip — and scores how demanding it is, from a one-liner to expert-level work.
02 — Route
The AI market gets shopped
Open benchmark scores set a quality bar the answer must clear. Among the models above it, live per-token pricing picks the right one for you. Hard tasks demand the frontier, easy ones don’t pay for it.
03 — Receipt
The answer comes priced
You get the full response, the exact cost in tokens and dollars, what the premium model would have charged, and an estimate of the electricity the request used — every time, on every plan.
Every frontier model. One account.
No tabs, no juggling AI subscriptions, no guessing which one is good at what.
Anthropic
Claude
Writing, long documents, and careful reasoning.
OpenAI
GPT & GPT Image
Broad capability, fast tiers for everyday work — and image generation.
Gemini, Nano Banana & Veo
Huge context windows, sharp translation, images and video.
xAI
Grok & Grok Imagine
Frontier reasoning at the sharpest price, and the cheapest images.
ByteDance
Seedance
Video by the second, at the lowest price per clip.
Eighteen models from five providers on every plan, routed live by price and benchmark quality. Plus adds the extended catalogue — DeepSeek, Qwen, Kimi and FLUX — for twenty-three models from nine providers.
Energy
See the estimated electricity behind every answer.
Nobody can tell you exactly what one AI request cost in watt-hours — batch size, active parameters and hardware generation all move the answer, and none of them are disclosed. So MPortal shows a range, computed from the tokens your request actually used and the model’s active parameter count, using the published EcoLogits method. Beside it: how that compares with every other model considered for the same request, and what each provider publishes about its own energy and carbon.
It covers operational electricity for that inference only. It excludes model training, hardware manufacturing, networking beyond the rack and your own device. It is not a carbon figure, not a total impact score, and not a grade — and we never rank a provider low for publishing nothing.
An estimate labelled as an estimate is worth more than a score that looks certain.
Budgets
A weekly AI budget that does the worrying for you.
When you sign up, MPortal asks you: how much do you plan to spend on AI each week? That number becomes your budget. Every request is priced against it, and as the week’s spend approaches the line, the router quietly shifts toward the most economical models that still get the job done — instead of cutting you off. You can also set a ceiling for any single request, and MPortal will refuse rather than exceed it.
Spend a number you chose, never a number that surprised you.
Stop overpaying for intelligence.
Start free. Keep the receipts.
The Solo plan costs nothing and asks for no card: unlimited routing decisions and estimates, plus about 10¢ of real usage every week. Upgrade only when you outgrow it.
Free at app.minott.tech · no card, no invite · talk to us about team and enterprise plans.