Make your first API request

Before you begin#

• Create a MomAPI key that is enabled and has not expired.
• Ensure your account balance and key quota permit the request.
• Install curl in your terminal. Start with a short prompt to check access and billing.

1. List available models#

Replace the placeholder and run this command. Copy a complete model id from the result for the MODEL_ID variable below.

export MOMAPI_API_KEY='YOUR_MOMAPI_API_KEY'

curl --fail-with-body 'https://api.momapi.link/v1/models' \
  --header "Authorization: Bearer ${MOMAPI_API_KEY}"

2. Send a streaming Responses request#

Replace MODEL_ID_FROM_YOUR_TOKEN with the full model ID from the previous step. Keep stream: true; some channels support only streaming Responses.

export MOMAPI_API_KEY='YOUR_MOMAPI_API_KEY'
export MODEL_ID='MODEL_ID_FROM_YOUR_TOKEN'

curl --fail-with-body --no-buffer \
  'https://api.momapi.link/v1/responses' \
  --header "Authorization: Bearer ${MOMAPI_API_KEY}" \
  --header 'Content-Type: application/json' \
  --data-binary @- <<JSON
{
  "model": "${MODEL_ID}",
  "input": "Reply with: Connection successful",
  "stream": true
}
JSON

3. Check the complete response#

A streaming response contains a sequence of SSE events. Check the ending event and any error as well as the text deltas. A successful Responses request should end in a completed state, such as response.completed.

If you see response.failed, response.incomplete, a disconnect, or a trailing error, retain the request ID and check usage logs. HTTP 200 or the first text delta alone does not prove successful completion.

Use an SDK or another client#

Use the same API key, Base URL, and full model ID. Your client must support the endpoint required by the model. A Chat Completions-only client cannot directly use a Responses-only channel.

For long tasks, keep reading the stream and configure an appropriate client timeout. Validate actual long requests with your chosen client and model.

Continue to client setup →

When a request fails#

Check the final URL, authorization header, model ID, endpoint format, and available quota. Then compare the error with usage logs. See troubleshooting for common cases.