Connect your first AI provider.

1. Create an account and connect a provider

Start your free trial, then open Providers & Models in your dashboard. Add your provider credentials and configure the models you want to use. Credentials are managed by your organization.

2. Create a Heimdall API key

Open API Keys and create a key for your application. Store it securely as the HEIMDALL_API_KEY environment variable. Use this key for Heimdall requests; keep your provider keys in the provider configuration.

3. List your available models

curl https://api.heimdall.run/v1/models \
  -H "Authorization: Bearer $HEIMDALL_API_KEY"

Copy an exact id from the response. Configured aliases appear by alias. Other models typically use provider/model-name. If the list is empty, check your provider credentials and model configuration.

4. Send a request and inspect the trace

Replace MODEL_ID_FROM_MODELS below with an ID returned in step 3.

curl https://api.heimdall.run/v1/chat/completions \
  -H "Authorization: Bearer $HEIMDALL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MODEL_ID_FROM_MODELS",
    "messages": [{"role": "user", "content": "Hello, Heimdall!"}]
  }'

Open Traces to inspect the model, status, token counts, latency, and secret-scan result.

Using an existing OpenAI-compatible client

Set its base URL to https://api.heimdall.run/v1, its API key to your Heimdall key, and its model to an ID from /v1/models. For streaming, use stream: true on the chat completions request. Support for optional parameters depends on the upstream provider.

Limits and troubleshooting

• 401: Check that you supplied a valid Heimdall API key.
• Model errors: Use an available model ID and verify provider access.
• 429: Check your request allowance and rate limits. Paid allowances renew monthly; the Free Trial does not renew.
• Data location: Heimdall hosts its gateway and stored traces in the EU. Your selected model providers have their own processing locations and terms.