OrcaRouter (OpenAI-compatible proxy) for Apify

Use OrcaRouter — an OpenAI-compatible LLM router that gives you one endpoint for models from OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen and more, plus an adaptive orcarouter/auto router — from inside the Apify platform.

This Actor runs in Standby mode as a small OpenAI-compatible HTTP proxy. You bring your own OrcaRouter API key (sk-orca-...); requests are forwarded to https://api.orcarouter.ai/v1 and billed to your OrcaRouter account. Apify only bills the Standby compute.

Disclosure: I'm an engineer on the OrcaRouter team.

Why use it

If you build Actors or AI agents on Apify, this gives you an OrcaRouter LLM backend that lives in the Apify ecosystem — a single OpenAI-compatible URL you can point any client or agent framework at, without writing your own proxy.

Get an OrcaRouter API key

Sign up at www.orcarouter.ai and create a key in the console. Keys start with sk-orca-.

Use it (Standby)

Start the Actor in Standby mode. You get a static URL, e.g. https://<your-standby-host>.apify.actor. The OpenAI-compatible base is that host + /api/v1.

Calling a Standby Actor requires your Apify API token (to authenticate to the Apify platform). You provide your OrcaRouter key in one of two ways:

Option A — key stored in the Actor input (recommended). Set orcaRouterApiKey in the Actor input. Then authenticate to Apify with your Apify token:

from openai import OpenAI

client = OpenAI(
    base_url="https://<your-standby-host>.apify.actor/api/v1",
    api_key="<YOUR_APIFY_TOKEN>",  # authenticates to Apify; the OrcaRouter key is read from the Actor input
)

resp = client.chat.completions.create(
    model="orcarouter/auto",
    messages=[{"role": "user", "content": "Hello from Apify!"}],
)
print(resp.choices[0].message.content)

Option B — bring the key per request. Pass your Apify token in the URL (?token=) and your OrcaRouter key as the OpenAI api_key:

client = OpenAI(
    base_url="https://<your-standby-host>.apify.actor/api/v1?token=<YOUR_APIFY_TOKEN>",
    api_key="sk-orca-...",  # forwarded upstream to OrcaRouter
)

You can also pass the OrcaRouter key explicitly with an X-OrcaRouter-Api-Key header.

Models

Use any model id OrcaRouter serves, namespaced by provider, plus the adaptive router:

Model id	Notes
orcarouter/auto	Adaptive router — picks an upstream per request (configurable in the OrcaRouter console)
openai/gpt-5.5	OpenAI flagship
anthropic/claude-opus-4.8	Anthropic flagship (reasoning model — do not send temperature)
google/gemini-3.5-flash	Google flagship
grok/grok-4.3, deepseek/deepseek-v4-pro, qwen/qwen3.7-max, ...	See the full catalog

Full catalog: www.orcarouter.ai/models. A bare model name (e.g. auto) is automatically namespaced to orcarouter/auto.

Supported endpoints mirror OrcaRouter: /chat/completions, /messages, /embeddings, /models.

Local development

npm install
npm test                       # unit + behavior tests (fake upstream, no key needed)
ORCAROUTER_API_KEY=sk-orca-... apify run   # run the Standby proxy locally
node scripts/orcarouter-live-test.mjs      # end-to-end live test against real OrcaRouter

Environment variables: ORCAROUTER_API_KEY (fallback key), ORCAROUTER_API_BASE_URL (override upstream, default https://api.orcarouter.ai/v1), ORCAROUTER_HTTP_REFERER / ORCAROUTER_X_TITLE (attribution).

License

Apache-2.0.