en-api-overview.md
OmniaKey exposes three protocol surfaces on one API host:
https://api.omniakey.com
Use the protocol your SDK or coding agent already speaks. The same OmniaKey API key works across the public Claude, GPT, Gemini, and Grok catalog.
| Protocol | Base URL | Primary use |
|---|---|---|
| OpenAI-compatible | https://api.omniakey.com/v1 | OpenAI SDKs, Cursor, Codex, Cline, aider |
| Anthropic-native | https://api.omniakey.com | Claude Code, Anthropic SDKs |
| Gemini-native | https://api.omniakey.com/v1beta | Google Gen AI SDKs |
The Anthropic SDK appends /v1/messages by itself. Use the bare host for Anthropic-native tools.
| Endpoint | Recommended models | Response protocol | Best for |
|---|---|---|---|
/v1/chat/completions | Claude, GPT, Gemini, Grok | OpenAI Chat Completions | Most tools and SDKs |
/v1/responses | GPT | OpenAI Responses | Codex-style GPT workflows |
/v1/messages | Claude | Anthropic Messages | Claude Code and Anthropic SDKs |
/v1beta/models/{model}:generateContent | Gemini | Gemini Generate Content | Gemini-native SDKs |
/v1/models | Claude, GPT, Gemini, Grok | OpenAI-compatible model list | Model discovery |
/v1beta/models | Gemini | Gemini model list | Gemini-native model discovery |
The current public documentation covers text and coding-agent workflows only:
Do not assume image generation, audio, embeddings, rerank, files, fine-tuning, or batches are supported unless they are added to this docs site and shown in the dashboard catalog.
OmniaKey does not silently substitute, quantize, or distill models. The model id you request is the model routed upstream. If a model is unavailable, the request fails instead of being hidden behind another model.
Requests spend from your prepaid account balance. Usage is metered per token and visible in the Dashboard. API keys can also have per-key spending caps.
API keys, headers, and SDK auth conventions
Chat Completions, Responses, and Models
Status codes, quota errors, and troubleshooting
⌘I