Using AI Models in n8n Workflows Without an International Card

The problem with OpenAI credentials in n8n

n8n's built-in OpenAI integration calls OpenAI directly — which requires an OpenAI API key billed in USD. For Indian users and teams who cannot use international cards, this blocks access to AI-powered automation.

The fix is to use n8n's custom base URL support. Any OpenAI-compatible API works as a drop-in, including AICredits, which accepts INR payments via Razorpay.

Setup: connect n8n to AICredits in 3 steps

Step 1: Go to Settings → Credentials → Add Credential → OpenAI API. Enter your AICredits sk- key in the API Key field. Expand Additional Fields and set Base URL to https://api.aicredits.in/v1. Save the credential.

Step 2: Add an OpenAI node to your workflow. Select your AICredits credential. In the Model field, type the model ID you want — for example openai/gpt-4o-mini for a fast cheap model or openai/gpt-4o for higher quality.

Step 3: Run the node. Your workflow is now using AICredits with INR billing.

AI Agent node: one critical setting

If you are using the AI Agent node with a Chat Model sub-node, there is one important setting to check. In the OpenAI Chat Model sub-node, find the option called Use Responses API and make sure it is turned OFF. When this is on, n8n sends requests to a different endpoint that AICredits does not support.

With Use Responses API off and your AICredits credential selected, the AI Agent node works with any model in the AICredits catalogue — including Claude Sonnet, Gemini Flash, DeepSeek, and Mistral alongside the OpenAI models.

Cost for common n8n automation patterns

Email classification workflow (read email, classify intent, route to channel):

• ~400 input tokens + 50 output tokens per email
• With GPT-4o Mini at ₹14/M input: ~₹0.006 per email
• ₹200 top-up processes ~33,000 emails

Summarisation workflow (2,000-token email → 200-token summary):

• With GPT-4o at ₹240/M input: ~₹0.53 per email
• ₹200 top-up processes ~380 emails

Document Q&A workflow (4,000-token context, 500-token answer):

• With Claude 3.5 Haiku at ₹96/M input: ~₹0.43 per request
• ₹500 top-up handles ~1,160 queries

For high-volume workflows, use GPT-4o Mini or Gemini Flash to keep costs low. Upgrade to GPT-4o or Claude Sonnet only for tasks where quality clearly matters.