n8n

The n8n OpenAI node is designed primarily around official OpenAI credentials. When connecting to Tokeness, the HTTP Request node is more straightforward and easier to debug. Get it working with HTTP Request first, then decide whether to wrap it into a credential or sub-workflow depending on your n8n version.

Connection Details

Field	Value
Method	POST
URL	https://n.tokeness.dev/v1/chat/completions
Authorization	Bearer your Tokeness API key
Content-Type	application/json
Model	The full model name from the Tokeness model catalog

1. Prepare Tokeness

Prepare the following first:

1. A dedicated API key for n8n.
2. A chat model name.
3. Available balance.

Name the key something like:

n8n-workflow

If you have multiple workflows in n8n, you can split keys by workflow.

2. Create an HTTP Request Node

Add an HTTP Request node to your n8n workflow.

Fill in:

Method: POST
URL: https://n.tokeness.dev/v1/chat/completions

You can leave Authentication set to None for now and add the header manually. Once it works, switch to Bearer Auth or Header Auth credentials.

3. Configure Headers

Add two headers:

Authorization: Bearer YOUR_TOKENESS_API_KEY
Content-Type: application/json

Do not put the real key directly into a workflow JSON that will be exported. For long-term use, store the key in n8n Credentials or an environment variable.

4. Configure the Body

Set the body type to JSON and fill in:

{
  "model": "YOUR_MODEL_NAME",
  "messages": [
    {
      "role": "user",
      "content": "Summarize in one sentence: {{$json.text}}"
    }
  ]
}

Replace YOUR_MODEL_NAME with the model name from the Tokeness model catalog.

If the upstream node does not have a text field, test with a fixed string first:

{
  "model": "YOUR_MODEL_NAME",
  "messages": [
    {
      "role": "user",
      "content": "Reply with ok only"
    }
  ]
}

5. Extract Text from the Response

The text returned by Chat Completions is typically at:

choices[0].message.content

You can read this field in a subsequent Set, Code, or IF node.

If you use the Responses API, change the URL to:

https://n.tokeness.dev/v1/responses

Example body:

{
  "model": "YOUR_MODEL_NAME",
  "input": "Reply with ok only"
}

The output structure of the Responses API differs from Chat Completions, so downstream nodes must read the actual returned fields.

6. Use Credentials

n8n HTTP Request credentials support Bearer Auth and Header Auth.

Bearer Auth:

Bearer Token: your Tokeness API key

Header Auth:

Name: Authorization
Value: Bearer your Tokeness API key

After creating the credential, select it in the HTTP Request node. This way the key is not exposed when the workflow is exported.

7. Workflow Tips

Scenario	Recommendation
Batch processing	Create a dedicated key for the workflow and set a quota
Long text	Set a reasonable timeout and split the input if needed
Retries	Limit retry count to avoid burning balance in a failure loop
Logging	Do not log full request headers
Concurrency	Control batch node concurrency to avoid sudden cost spikes

8. Troubleshooting

Symptom	Action
401 Unauthorized	The Authorization header should be Bearer KEY
404 Not Found	The URL should be https://n.tokeness.dev/v1/chat/completions
model not found	Re-copy the model name from the Tokeness model catalog
JSON parse error	Set the body type to JSON and check quotes and commas
Expression is empty	Test with fixed text first, then connect upstream fields
No Tokeness logs	The request did not reach Tokeness — check the URL and whether the node executed
Workflow calls repeatedly	Check loop nodes, retry settings, and error branches

External Documentation

• n8n HTTP Request Credentials
• n8n OpenAI Node
• n8n OpenAI Credentials