API Access overview
API Access lets you call Steinkauz AI from your own applications using OpenAI-compatible HTTP endpoints and familiar SDKs. The Platform governs customer-owned applications through tenant-scoped API access.
Steinkauz AI is not affiliated with OpenAI. Compatibility is limited to the endpoints documented in this section.
API keys and usage belong to your active organization. Create keys in API access while the organization you want to integrate with is selected.
The API authenticates the key, applies optional Budgets, and evaluates the organization’s D×E routing matrix before sending input to a provider. Requests fail if required evidence cannot be recorded. Client-defined tools are executed by your application; Steinkauz records reported calls and results as evidence.
Supported endpoints
| Endpoint | Method | Purpose |
|---|---|---|
/v1/models | GET | List models available to your context and API key |
/v1/models/{model} | GET | Retrieve one listed model |
/v1/chat/completions | POST | Chat completions (streaming and non-streaming) |
/v1/responses | POST | Responses API core (streaming and non-streaming) |
/v1/embeddings | POST | Create embeddings on embedding-capable providers |
/v1/audio/transcriptions | POST | OpenAI-compatible multipart audio transcription |
/v1/decisions | POST | Predicate, choice and score inference with native decision models |
These endpoints implement the commonly used interoperable subset, not every OpenAI product API. Stored Responses and supported continuation are available as documented in Responses. Assistants, Files, Batches and image-generation endpoints are not offered. File bodies are not retained.
Base URL
Use your Steinkauz AI platform origin as the API base:
https://platform.steinkauz.ai/v1Authentication
Requests require a Steinkauz AI API key in the Authorization header. Choose a complete model identifier from /v1/models. See Authentication.
Billing & limits
Steinkauz AI does not charge per inference request. You pay upstream providers directly. BYOK stays uncapped by Steinkauz AI until you set optional Budgets. Usage is recorded in Activity with API key attribution.
Optional BYOK Budgets remain. There is no included inference credit.
When Remaining reaches zero for a configured Budget unit, inference requests return 402 with insufficient_quota on the next gated attempt. See Usage Budgets and Usage, costs, and activity details.
Rate limits
Request-rate limits apply per API key, and concurrent streaming caps apply per creating user across that user’s keys and organizations. See Errors & limits.
API routing policy
Each API key has an API routing policy: an E-class floor plus a D-class default and floor. Inference uses the tenant’s D×E matrix. Ordinary model listing filters by the provider E-floor; every concrete inference request checks its effective D-class against the matrix.
Who can create API keys
Members can create and revoke their own inherited-budget keys. Only owners and admins can create or manage dedicated-budget keys. See Members, roles & invites.
Usage and activity
Allowed and blocked decisions appear in Activity → Requests for the active organization. Inspect attribution, policy, provider/model, recorded usage and evidence correlations together.
Next steps
- Create an API key in the platform app (API access)
- Authentication
- Chat completions, Responses, Embeddings and Audio transcription