Skip to content
ALTAIRA LABS
DocsBlog
OmniaOmnia

Configure OpenRouter provider

Point an Omnia Provider at OpenRouter so an agent can use any model it hosts with one API key.

OpenRouter speaks the OpenAI wire format, so the Provider uses type: openai with spec.baseURL set to OpenRouter and spec.headers carrying its attribution headers.

Prerequisites

Section titled “Prerequisites”
  • An OpenRouter API key, created at openrouter.ai/keys.
  • The Omnia operator installed in the cluster.

1. Create a secret for the API key

Section titled “1. Create a secret for the API key”

The Secret must be in the same namespace as the Provider.

Terminal window
kubectl create secret generic openrouter-credentials \
  --namespace agents \
  --from-literal=OPENAI_API_KEY='sk-or-v1-...'

OPENAI_API_KEY is the key the controller looks for by default on a type: openai Provider. To use another key name, set spec.credential.secretRef.key.

2. Create the provider

Section titled “2. Create the provider”
apiVersion: omnia.altairalabs.ai/v1alpha1
kind: Provider
metadata:
  name: openrouter
  namespace: agents
spec:
  type: openai
  model: anthropic/claude-sonnet-4                # OpenRouter model ID
  baseURL: https://openrouter.ai/api/v1


  headers:
    HTTP-Referer: https://your-app.example.com    # attribution; replace with your app's URL
    X-Title: omnia                                # name shown in OpenRouter analytics


  credential:
    secretRef:
      name: openrouter-credentials


  capabilities:
    - text
    - streaming
    - tools
    - json

The model IDs here are illustrative. Take the exact ID from openrouter.ai/models; it has the form <vendor>/<model>.

Do not put Authorization or any other header the OpenAI provider sets itself in spec.headers. The request fails with custom header "Authorization" collides with built-in header set by provider.

3. Set pricing for non-OpenAI models

Section titled “3. Set pricing for non-OpenAI models”

Cost tracking uses spec.pricing when set. Without it, PromptKit prices only models in its built-in OpenAI table, after stripping the vendor prefix: openai/gpt-4o is priced as gpt-4o, but anthropic/..., google/... and meta-llama/... models are unpriced. Set the rates from OpenRouter’s model page:

spec:
  pricing:
    inputCostPer1K: "0.003"    # illustrative; use OpenRouter's current rates
    outputCostPer1K: "0.015"

4. Use it from an AgentRuntime

Section titled “4. Use it from an AgentRuntime”
apiVersion: omnia.altairalabs.ai/v1alpha1
kind: AgentRuntime
metadata:
  name: my-agent
  namespace: agents
spec:
  promptPackRef:
    name: my-pack
  providers:
    - name: default
      providerRef:
        name: openrouter
  facades:
    - type: websocket

Verify

Section titled “Verify”
Terminal window
kubectl get provider openrouter -n agents \
  -o jsonpath='{range .status.conditions[*]}{.type}={.status} {.reason}{"\n"}{end}'

SecretFound and CredentialConfigured should be True. EndpointReachable should also be True: the controller probes baseURL without credentials, and any HTTP response, including a 401, counts as reachable. False means the connection itself failed (DNS, timeout or refused), and the Provider phase becomes Unavailable. CredentialValid reports a request to baseURL + /models (https://openrouter.ai/api/v1/models) sent with the key.

Troubleshooting

Section titled “Troubleshooting”

401 Unauthorized on every request

Section titled “401 Unauthorized on every request”

Check that the secret holds an OpenRouter key (sk-or-v1-...), not an OpenAI key. Check the first characters without printing the whole key:

Terminal window
kubectl get secret openrouter-credentials -n agents \
  -o jsonpath='{.data.OPENAI_API_KEY}' | base64 -d | cut -c1-9

Requests succeed but OpenRouter does not attribute the app

Section titled “Requests succeed but OpenRouter does not attribute the app”

Check that HTTP-Referer and X-Title are set in spec.headers.

Section titled “Related pages”