Point the client at the Clyvel Gateway
Use the Clyvel Gateway base URL instead of calling the provider endpoint directly. Keep the application-facing request shape compatible with the SDK behavior your workload depends on.
All integrations
OPENAI SDK INTEGRATION
OpenAI SDK Gateway Integration for Production AI
Clyvel is designed to preserve the OpenAI-compatible request pattern while moving production traffic through one operational layer. Teams can keep a familiar client interface while gaining request-level telemetry, cost attribution, routing context and provider isolation around the model call.
WHEN THIS FITS
A practical production use case.
Teams already using an OpenAI-compatible client and wanting a controlled production request path.
Applications that need model, token, latency, status and estimated cost evidence per request.
Organizations that want provider credentials separated from application-facing gateway credentials.
CONNECTION PATH
Connect the model request layer, then verify evidence.
Use a Clyvel workload credential
Authenticate the application to Clyvel with the scoped key created for that workspace or workload. Provider credentials remain managed behind the operational layer.
Send one real request and verify telemetry
Confirm model, provider, status, latency, token usage and cost evidence in Requests and Overview before moving additional production traffic.
01
One observable request path for OpenAI-compatible traffic.
02
Request-level latency, tokens, status and cost evidence.
03
A clean migration path toward routing, reliability and governance controls.
FAQ
Integration questions.
Do I need to rewrite my OpenAI client to use Clyvel?
The integration is designed around an OpenAI-compatible request path, so the main application change is typically the gateway base URL and credential strategy. Validate the exact endpoints and streaming behavior your workload uses before migrating production traffic.
Does Clyvel replace my model provider account?
No. Your provider account remains separate. Clyvel becomes the operational request layer used to route, observe and control production model traffic.
NEXT STEP
Connect one workload before expanding the rollout.
Start with a controlled request, confirm that operational evidence is trustworthy, and only then move more production traffic or add routing and reliability policy.