TYO Prism: One Endpoint, Every AI Model
If you build with AI, you know the dilemma. The strongest models are brilliant — and expensive. The cheap ones are fine for most everyday requests — until they aren't. So you either overpay by sending everything to a top-tier model, or you burn engineering time wiring up provider after provider and deciding, request by request, which one to call.
TYO Prism does that deciding for you. Point your app at one endpoint, plug in your own API keys, and every request takes the smart path: easy work goes to your cheap models, hard work goes to your strong ones — and a live dashboard shows you exactly what you saved.
Your keys, not ours
Prism is bring-your-own-key. You add the API keys you already have — DeepSeek, OpenAI, Anthropic (Claude), Google (Gemini), Qwen, Kimi, GLM — or point Prism at your own self-hosted models (vLLM, Ollama, anything with an OpenAI-compatible URL). Tokens are billed by your providers at your prices; Prism never marks them up and never resells them.
Your provider keys are stored encrypted and are never shown again after you save them. Your prompts and responses are never stored — Prism keeps only the metadata it needs to run your dashboard: token counts, costs and latency. The only party that reads your prompt is the model provider Prism routes it to.
How it works
Your app already speaks Prism's language. Anything that works with the OpenAI API works with Prism — change one line and you're routed:
export OPENAI_BASE_URL=https://prism-api.tyo.com.au/v1
export OPENAI_API_KEY=<your Prism key>
Then pick how each request is routed, simply by the model name you send:
auto — the smart path. Prism sizes up each request and sends it to the
cheapest of your models that can handle it well, escalating the hard ones to your
strong tier. Requests that need special abilities — tools, vision, very long
context — are only ever sent to models that have them.
cascade — your ladder, your order. Number your models on the dashboard —
layer 1 first, then 2, then 3 — and Prism tries them strictly in your order,
stepping up a layer whenever one fails or is overloaded. Start with the cheapest
model you trust and let the ladder carry the rest.
Or name a model. Send an exact upstream name and Prism routes straight to it — no guessing, when you know exactly what you want.
Streaming is preserved end to end, and if a provider is down or rate-limited, Prism steps to your next option automatically instead of failing your request.
See what you saved
Every request is measured against what it would have cost on your most expensive model. The dashboard shows your requests, tokens, spend and — the number that matters — savings, live and by the day. Routing research consistently finds that most everyday requests don't need a frontier model; Prism turns that finding into money you keep.
Plans
Your request allowance resets on the 1st of each month, and the dashboard always shows where you stand. Prism charges a flat subscription for the routing service — never a percentage, never a markup on tokens.
Getting started
1. Sign in free at prism.tyo.com.au with
your TYO account — no card. 2. Add an upstream: pick a provider preset, paste
your API key (stored encrypted, shown never again). 3. Point your app at
https://prism-api.tyo.com.au/v1 with your Prism key and send
model: "auto". That's it — watch the request land on your dashboard.
Prism works with every OpenAI SDK and tool that lets you set a base URL — Python, Node, curl, LangChain, coding agents, chat UIs. The quickstart guide has copy-paste examples.
Ready to route smarter? Start free at prism.tyo.com.au — 1,000 requests a month on your own keys, no card, no commitment. Questions or feedback? Get in touch.