A drop-in OpenAI-compatible gateway to every AI
If your code already speaks the OpenAI API, TDSPRO is a drop-in: change the base URL to our /v1 endpoint, keep your existing OpenAI SDK, and reach 20+ models through one key. No rewrites, no new client libraries.
Truly drop-in
Same request and response shapes as the OpenAI API. chat/completions, embeddings, model list — all there.
One key, many models
Reach 20+ models behind a single key and a single endpoint — pick a model by name or by alias.
Fallback built in
If a model is busy or down, the gateway retries on an alternative in milliseconds — your app keeps responding.
Cache that cuts the bill
Identical requests are served from cache and return the same model — honest savings, no silent downgrades.
Why "compatible" matters
OpenAI-compatible means you do not learn a new API. The endpoint paths, the JSON you send, and the JSON you get back match what your tools already expect. That is the difference between a five-minute switch and a multi-day migration.
Point any OpenAI SDK — Python, Node, or a CLI — at our base URL, pass your TDSPRO key as the bearer token, and you are calling 20+ models immediately. Aliases like fast, smart, code, vision and chat let you switch model class with a single string.
What you keep — and what you gain
You keep your code, your SDK, and your mental model of the API. You gain one key instead of many, automatic fallback, a token-saving cache, and one dashboard for budgets and usage across every model.
Start free — no card required
Create a free key and get 20+ models on the house. Crypto is only for upgrades — never to begin.
The free tier is genuinely free: 50 requests/day across all models, OpenAI-compatible, no card. Upgrade only when you outgrow it.
Frequently asked
Is it really OpenAI-compatible?+
Yes. The /v1 endpoint mirrors the OpenAI API surface — chat/completions, embeddings and model listing — so existing OpenAI SDKs work unchanged. You only swap the base URL and the key.
Do I need a new client library?+
No. Use the OpenAI SDK you already have, or plain HTTP. Set the base URL to our /v1 endpoint and go.
Will the model name be honest?+
Always. Every response carries the real model name. Caching returns the same model — we never silently downgrade.
How do I start?+
Create a free key (50 requests/day, no card), copy the base URL from the connect page, and make your first call.