Documentation
CostSlice docs
CostSlice is a pass-through OpenAI proxy plus a tenant P&L. It is not an AI gateway: no routing, fallbacks, semantic cache, or virtual keys. We store usage metadata only — never prompts or completions.
- 10-minute quickstart — Python, Node, and Vercel AI SDK
- Proxy host:
https://api.costslice.com/v1 - App:
https://costslice.com - Usage API:
GET https://costslice.com/api/v1/spendwithx-cs-key
What we store
Per request: org, request id, timestamp, provider, model, tenant, feature, x-cs-env (prod or staging), token counts, cost, latency, status, and whether cost came from the provider, an estimate, or an unpriced model. No body. We do not log payloads. We do not train on customer data because we do not keep it.
Headers
| Authorization | Your OpenAI API key. Forwarded. Never logged. |
| x-cs-key | CostSlice workspace key. Identifies your org. |
| x-cs-tenant | Opaque tenant id, max 128 chars. Not an email. |
| x-cs-feature | Feature name, max 128 chars. |
| x-cs-env | prod | staging | dev. Defaults to prod. |
Missing tenant: if x-cs-tenant is absent we use OpenAI user as the tenant id, then fall back to untagged. A missing tenant ID still completes and lands in untagged. Starter at 10,000 requests: New proxy calls stop until you upgrade. That is the Starter plan cap, not a per-customer block.