PRODUCT
One endpoint, every model: how Qevron, our OpenAI-compatible gateway, works
2 min read
Point your code at a single OpenAI-compatible API and reach our in-house models plus 43+ providers — with routing, caching, monitoring and cost analytics in one place.
Most teams don’t have a model problem; they have an integration problem. Every provider has its own SDK, its own quirks, its own bill. Qevron is our answer: one gateway in front of all of it.
The OpenAI-compatible interface, on purpose
Since OpenAI shipped the Chat Completions API in 2023, it has become the de-facto interface of the industry — Google, Anthropic, AWS Bedrock and every open-weight inference server now speak some form of it. Qevron exposes that same OpenAI-compatible surface, so your existing code points at one endpoint and reaches our first-party models alongside 43+ external providers by changing little more than a base URL.
What the gateway does beyond proxying
A gateway earns its place by doing the work you would otherwise rebuild in every app:
- Routing across first-party and external models, with fallback when a provider is slow or down
- Caching to cut repeated cost and latency
- Live monitoring and cost analytics, so you can see every token and where the spend goes
- A built-in tokenizer, multi-model orchestration and MCP tools
- One place to enforce keys, budgets and policy instead of scattering them across services
Why this is also a sovereignty decision
Standardising on one OpenAI-compatible gateway is portability insurance. Models get deprecated, providers change prices, regions matter. When the interface is uniform, swapping or adding a model is a configuration change, not an application rewrite. And because Qevron fronts our own models too, you can keep sensitive traffic on infrastructure you control.
A gateway is not magic, and we won’t pretend otherwise. “OpenAI-compatible” is a de-facto standard, not full feature parity — some provider-specific features live only in native APIs — and any gateway adds a network hop and a place keys pass through. The trade is deliberate: you accept one well-run layer to avoid a stack of brittle ones.
Qevron is the “our gateway” in “Our models. Our gateway. Your AI.”
Qevron is the control plane the rest of the platform is built on; every other Arpanet product reaches its intelligence through it.