Free AI Gateway quota: a private, OpenAI-compatible router at zero cost

We opened 50 free seats on ekxo Gateway — a private, OpenAI-compatible AI router with real quota attached and no card required. If you’ve been comparing gateways, or looking for the cheapest way to run inference without shipping your data to someone else’s cloud, this is the one to try first.

What you get, free

500,000 tokens every month — enough to build and ship something real, not just a demo.
15 requests / minute, 10,000 tokens / minute.
Sovereign models on our own GPUs — Google for chat completion and Black Forest Labs for image generation, run as private inference on infrastructure we operate. Your prompts and outputs stay with us, not a third-party API.
One OpenAI-compatible endpoint + API keys — change one base URL and you’re live. Keep your existing SDK.
Built-in router, cache, and observability — the middleware you’d otherwise wire up yourself.
Free. One per account, no card at checkout.

Only 50 seats are open at a time. Keep yours by using it — a seat with no usage for 30 days is released back to the pool — so claim one while it’s here.

For engineering leaders: own the layer, control the spend

A raw API key gives you a model. A gateway gives you the layer in front of it — where cost, privacy, and reliability are actually decided. Every request is governed in one place: predictable caps instead of surprise bills, private inference instead of data leaving your stack, and one accountable endpoint instead of a tangle of provider SDKs. When the best model changes next quarter — and it will — you switch behind the interface, not across your whole codebase.

For developers: drop-in, not a rewrite

It speaks the OpenAI API. Point your base_url at the gateway, drop in your key, and your existing client just works — Python, TypeScript, anything that already talks to OpenAI. Routing, caching, and request logs come turned on, so you spend your time on the feature, not the plumbing.

The best AI router is the one you can hold accountable

One endpoint, every model. Bring your own provider keys, or use our optimized open models running on our own vLLM — private by default. Router, cache, and observability are built in, and frontier-model routing through a single unified gateway is on the roadmap. The reach of an aggregator, on infrastructure you can actually hold accountable.

Claim a free seat →

FAQ

Is the AI Gateway really free? Yes — the Free tier is $0 with no card at checkout. There are 50 seats, and a seat is released back to the pool if it goes 30 days without any usage.

How much free AI quota do I get? 500,000 tokens per month, at 15 requests/minute and 10,000 tokens/minute. Need more? Lite, Pro, and Max scale the same gateway up to 100M tokens/month.

What is an AI gateway or router? A single OpenAI-compatible endpoint that sits in front of your models and governs every request — routing, caching, key management, and observability — so cost, privacy, and reliability are controlled in one place instead of per-provider.

Which models can I call? Google for chat completion and Black Forest Labs for image generation — both private, sovereign inference on our own GPUs — with your own provider keys supported through the same endpoint. Unified frontier-model routing is coming.

Do I need a credit card? No. The Free tier requires no card. Paid tiers keep a card on file for the next period.

Is my data private? Yes. Sovereign models run on infrastructure we operate — your prompts and outputs don’t leave for a third-party API.