Teams running more than one AI provider, whether for different models on different workloads or a fallback for reliability, end up maintaining separate auth, separate logging, and separate cost tracking for each one. Observra replaces that with a single gateway URL in front of every provider.
How routing works
The provider is identified from the gateway URL itself. There is no auto-detection and no ambiguity. Each provider has a dedicated adapter, so requests are forwarded with the correct shape and headers for that provider. Your application keeps its existing SDK; only the base URL and gateway key change per environment.
Supported providers
- OpenAI
- Anthropic
- Gemini
- Groq
- Azure OpenAI
- Ollama
- OpenRouter
One observability schema across providers
Every request through the gateway produces the same structured Observation: latency, tokens, cost, and errors in one schema, whichever provider handled it. Comparing two providers, or moving a workload from one to the other, does not mean rebuilding your monitoring.