Get started

Your SDK, connected to Renio.

Keep your provider’s formats. Set your SDK’s URL and authentication to use Renio with a re_ access key, following the examples below. The vault then supplies the real provider key for you.

The rule, in one sentence

Renio identifies you only through your re_ access key in the Authorization: Bearer header. For OpenAI, Mistral and OpenRouter, this is already the native header: set your SDK’s key to your re_ key. For Anthropic and Gemini (which normally use x-api-key / x-goog-api-key), explicitly set the Authorization header — the proxy adds the native authentication expected by the provider.

In the examples below, re_xxxxxxxxxxxxxxxxxxxx is a placeholder — replace it with the secret of your access key, shown only once when it is created.

OpenAI

https://api.renio.ai/openai/v1/chat/completions

curl https://api.renio.ai/openai/v1/chat/completions \
  -H "Authorization: Bearer re_xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

SDK — OpenAI SDK: set api_key to your re_ key and base_url to the URL above (without the endpoint suffix).

Anthropic (Claude)

https://api.renio.ai/anthropic/v1/messages

curl https://api.renio.ai/anthropic/v1/messages \
  -H "Authorization: Bearer re_xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-sonnet-5","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

SDK — Anthropic SDK: keys normally go in x-api-key; instead, send your re_ as Authorization: Bearer (using the SDK’s authToken option or defaultHeaders). The proxy sets x-api-key and anthropic-version for you.

Mistral

https://api.renio.ai/mistral/v1/chat/completions

curl https://api.renio.ai/mistral/v1/chat/completions \
  -H "Authorization: Bearer re_xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"mistral-large-latest","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

SDK — Native Bearer authentication: set the SDK key to your re_ key.

OpenRouter

https://api.renio.ai/openrouter/api/v1/chat/completions

curl https://api.renio.ai/openrouter/api/v1/chat/completions \
  -H "Authorization: Bearer re_xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-4o","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

SDK — Native Bearer authentication: set the SDK key to your re_ key. The exact cost returned by OpenRouter is used for your budgets.

Gemini (Google)

https://api.renio.ai/gemini/v1beta/models/gemini-2.5-pro:generateContent

curl https://api.renio.ai/gemini/v1beta/models/gemini-2.5-pro:generateContent \
  -H "Authorization: Bearer re_xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"role":"user","parts":[{"text":"Hello"}]}],"generationConfig":{"maxOutputTokens":256}}'

SDK — Gemini keys normally go in x-goog-api-key or ?key=; instead, send your re_ as Authorization: Bearer (the proxy removes any ?key= from the URL and sets x-goog-api-key for you).

What you should know

Capped access keys — Include an output limit in every generation: max_tokens, max_completion_tokens, max_output_tokens or generationConfig.maxOutputTokens, depending on your provider. The examples above include one. Unpriced models, multiple simultaneous generations or remote content whose cost cannot be estimated are refused with HTTP 422 and budget_measurement_unavailable. This check applies to access keys with a spending cap.
Unavailable budget counter — New controlled requests are refused with HTTP 503 and budget_unavailable. A request already sent to the provider may still incur charges, even if its stream is interrupted.
Mistral streaming — Usage may be absent from the stream; in that case the cost is marked “unmeasured” rather than invented.
OpenRouter — The exact cost returned by OpenRouter is used for your budgets. Interrupting a stream does not always stop upstream billing.
Images, audio and hosted tools — Their cost cannot be estimated reliably enough for capped access keys. For uncapped access keys, measurement depends on the adapter; for example, text pricing may be applied to Gemini audio usage. The journal does not replace the provider’s invoice.