Usage Control · BYOK
Bring your own key — Shanc brings the context
Shanc serves the context; inference runs on your own API keys. Plug in Anthropic, OpenAI, or your cloud provider’s endpoints, and every model call happens on your accounts — your negotiated rates, your data agreements, your choice of model, swappable as the market moves.
- Works with your API keys and cloud endpoints — Anthropic, OpenAI, Bedrock, Azure OpenAI, Vertex
- Inference under the data agreements you’ve already negotiated
- Swap or mix models without touching the context layer
How it works
Your keys
Model calls run on your accounts
Connect your API keys or your cloud provider’s AI endpoints, and that’s where inference happens. Shanc’s job stays serving definitions; which model reasons over them is your call, made on your infrastructure.
Your terms
The agreements you negotiated still hold
Your DPAs, retention terms and residency requirements travel with your keys — the model relationship stays between you and your provider, on the contract your legal team already approved. Adopting Shanc adds a context layer, not a new place your data goes for inference.
Your choice, ongoing
Change models without changing anything else
The context layer is model-agnostic: definitions served to Claude today work identically with whatever you route to next quarter. When a better or cheaper model ships, switching is a key swap — the layer, and everything built on it, stays put.
Your spend, your bill
Inference costs land where you can see them — on your own invoice.
With BYOK, model usage runs at the rates you’ve negotiated and counts toward the cloud commitments you already have. Shanc’s pricing covers the layer — sources connected and queries served — while inference stays a line on your own bill, controlled with your own limits.
Nobody’s betting on one model anymore
81% of enterprise CIOs expect to rely on two or more LLM providers in 2026. Infrastructure that assumes one model — or holds your keys for you — is built against where the market is going. (Dataiku / Harris Poll, survey of 600 CIOs)
55% of CIOs have already switched LLM providers at least once, mainly to reduce costs. The switch should cost a key swap, with your context layer unmoved. (Dataiku / Harris Poll)
57% of data teams report increased warehouse and compute spend, against 36% reporting increased budgets. Inference on your own negotiated rates and commitments is the difference between spend you manage and spend that happens to you. (dbt Labs, State of Analytics Engineering 2026)
The model is a choice you’ll make many times. The context layer is the part that shouldn’t move.