n8n
Build an n8n workflow on PromptsForLess with an OpenAI credential and custom base URL, or a plain HTTP Request node.
n8n's OpenAI Chat Model node works with PromptsForLess over Chat Completions. Leave the node's Use Responses API toggle and hosted tools off.
Add a credential
In n8n, create an OpenAI credential and enter your PromptsForLess key from the dashboard. Set Base URL to https://api.promptsforless.com/v1. Leave Organization ID empty.
Some n8n versions keep Base URL in the node's options instead of the credential. Set it there if so. If your version shows neither, use the HTTP Request node below so your key only goes to PromptsForLess.
Build a small chat workflow
- Add a Manual Trigger and a Basic LLM Chain.
- Attach an OpenAI Chat Model with your PromptsForLess credential.
- Pick a model from the list, or enter an exact ID from the catalog, such as
deepseek-v4.1-flash. - Keep Use Responses API off and leave hosted tools off.
- Send a fixed prompt such as “Write one friendly greeting” and run the workflow.
n8n's OpenAI Chat Model documentation describes the node options. For an AI Agent node, pick a model with tool calling.
Use an HTTP Request node
A server-side HTTP Request node gives you full control over the request:
| Field | Value |
|---|---|
| Method | POST |
| URL | https://api.promptsforless.com/v1/chat/completions |
| Authentication | Header Auth credential with Authorization: Bearer YOUR_PFL_KEY |
| Content type | JSON |
Keep the key in an n8n credential so it never lands in exported workflow JSON. Set the body to:
{
"model": "deepseek-v4.1-flash",
"messages": [{"role": "user", "content": "Write one friendly greeting."}]
}The reply is in choices[0].message.content, and usage shows the tokens used.
Troubleshooting
| Symptom | Fix |
|---|---|
| Model not in the list | Enter the exact model ID where the node accepts a custom value. |
/responses errors | Turn off Use Responses API. |
| Credential check fails | Check the key and base URL, then test with the HTTP Request node. |
| Workflow spends more than expected | Check retries, loops, conversation history and the number of input items. |
Compare model prices and use the quickstart for HTTP checks.