continuedev/continue · warning

[Continue] Azure endpoint detected but API key appears to be

Error message

[Continue] Azure endpoint detected but API key appears to be a standard Anthropic key (sk-ant-*). Azure endpoints require Azure API keys from your AI Foundry resource.

What it means

getAnthropicHeaders detects an Azure endpoint (via isAzureAnthropicEndpoint on the apiBase) but the supplied API key starts with sk-ant-, the prefix of standard Anthropic console keys. Azure AI Foundry endpoints authenticate with Azure-provisioned keys, so the request will fail auth; this warning tells you the key type does not match the endpoint type.

Source

Thrown at packages/openai-adapters/src/apis/AnthropicUtils.ts:75

      hostname.endsWith(".services.ai.azure.com") ||
      hostname.endsWith(".cognitiveservices.azure.com")
    );
  } catch {
    // Invalid URL - fall back to false
    return false;
  }
}

export function getAnthropicHeaders(
  apiKey: string,
  enableCaching: boolean,
  apiBase?: string,
): Record<string, string> {
  const isAzure = isAzureAnthropicEndpoint(apiBase);

  // Warn if Azure endpoint is used with Anthropic-style key
  if (isAzure && apiKey.startsWith("sk-ant-")) {
    console.warn(
      "[Continue] Azure endpoint detected but API key appears to be a standard Anthropic key (sk-ant-*). " +
        "Azure endpoints require Azure API keys from your AI Foundry resource.",
    );
  }

  const authHeaderName =
    isAzure && apiBase?.toLowerCase().includes("cognitiveservices.azure.com")
      ? "api-key"
      : "x-api-key";

  const headers: Record<string, string> = {
    "Content-Type": "application/json",
    Accept: "application/json",
    "anthropic-version": "2023-06-01",
    [authHeaderName]: apiKey,
  };

  if (enableCaching) {

View on GitHub (pinned to 5522c6f44c)

Solutions

1. Generate an API key in your Azure AI Foundry / Azure AI Services resource (Keys and Endpoint blade) and use that instead
2. Double-check apiBase: if you actually intend to hit api.anthropic.com, remove the Azure endpoint URL
3. Keep separate named profiles for Anthropic-direct and Azure so keys are not crossed

Example fix

# before
apiBase: https://my-resource.services.ai.azure.com/models
apiKey: sk-ant-api03-xxxx

# after
apiBase: https://my-resource.services.ai.azure.com/models
apiKey: <Azure Foundry key>

if (apiBase.includes("azure.com") && apiKey.startsWith("sk-ant-")) {
  throw new Error("Azure endpoint requires an Azure Foundry API key, not an sk-ant- key");
}

When it happens

Trigger: Configuring apiBase to an Azure AI Foundry / Models-as-a-Service Anthropic endpoint while keeping an Anthropic console key (sk-ant-...) from a non-Azure config.

Common situations: Migrating an existing Anthropic config to Azure and forgetting to swap the key; mixing credentials across profiles; copy-pasting a default Anthropic key into an Azure deployment's config.

Related errors

• Failed to fetch Anthropic models: ${response.status}
• Model ${modelName} does not have an apiKeyLocation or envSec
• AskSage adapter: missing apiKey. Provide it in your configur
• `env.deployment` is a required configuration property for Az
• `env.apiVersion` is a required configuration property for Az

AI-assisted analysis of continuedev/continue@5522c6f44c (2026-08-27). Data as JSON: /api/errors/4dc5bf8b6eedcdc4. Report an issue: GitHub.