FAQ

Frequently asked questions about Synthorai.

Which SDKs does Synthorai support?

The OpenAI SDK (via /v1/chat/completions and /v1/responses) and the Anthropic SDK (via /v1/messages). Set baseURL/base_url to your gateway and use your gateway API key — no other code changes.

Which base URL do I paste into my tool?

Two rules cover it: OpenAI-style tools (Codex, Cursor, Continue, Aider, LangChain ChatOpenAI, raw OpenAI SDK) want the origin WITH /v1 appended. Anthropic-style tools (Claude Code, Anthropic SDK, LangChain ChatAnthropic) want the origin WITHOUT /v1 — the SDK adds /v1/messages itself. The Tool Compatibility page has a full table. See page →

I set the env vars but my tool still connects to the original provider. What did I miss?

New environment variables only apply to shells opened AFTER you edited your profile. Open a fresh terminal, or run source ~/.zshrc / source ~/.bashrc in the current one. Verify with echo $OPENAI_BASE_URL or echo $ANTHROPIC_BASE_URL.

How do I know which model names are available to me?

Run curl https://synthorai.io/v1/models -H "Authorization: Bearer YOUR_KEY", or visit Console → Model Pricing. Pass the id column value as the model parameter in your request — the gateway handles routing automatically.

Can I use the Anthropic SDK with non-Claude models?

No. /v1/messages only routes to Claude models via configured Anthropic or Bedrock-Anthropic channels. For other providers use /v1/chat/completions, which works with every routed model.

Which protocol should I pick for a new project?

Use /v1/chat/completions for maximum compatibility — it works with every provider. Use /v1/responses for OpenAI tool-use and reasoning. Use /v1/messages if you are standardizing on the Anthropic SDK.

What happens when my quota runs out?

Requests will return a 403 status code with error type "quota_exceeded". You can top up in Console → Limits & Billing.

How is usage calculated?

Usage is tracked in tokens. Input and output tokens are counted separately and reported in the usage field of each API response.

Can I set per-user quotas?

Yes. Admins can configure per-user token quotas from Console → Members.

Are requests logged?

Yes. All requests are logged with model, tokens, latency, and status. Logs are retained for 30 days and available in Console → Logs.

Comparing spend across models? The model price comparison lists every routable model's per-token pricing, and the LLM API cost calculator turns your token volumes into a monthly bill estimate.