← Pricing & setup guides

n8n with an OpenAI-compatible API: connect a monthly model budget

Keep your n8n workflow. Connect Standard Compute for model requests, with a fixed monthly API budget and a small manual test before turning on automation.

The buying decision

Start with a manual HTTP request to verify your account and endpoint. Then use the OpenAI Chat Model node where its protocol and model options fit your workflow.

Import a workflow that makes one request

The download contains a manual trigger and one HTTP Request node. It has no API key, schedule or external action beyond the model request. Import the JSON through n8n’s workflow import menu, then configure the request node’s Header Auth credential before running it.

  1. Create a Header Auth credential: Name is Authorization; Value is Bearer followed by a space and your own Standard Compute API key.
  2. Select that credential in the Get a model response node. Keep it in n8n’s credential store.
  3. Execute the workflow manually. Inspect choices[0].message.content in the JSON response and check your Standard Compute dashboard allowance.

Use the Chat Completions endpoint

The starter sends POST https://api.stdcmpt.com/v1/chat/completions with model standardcompute, one short user message and max_tokens 128. It is a text-connection example. It does not establish support for image generation, embeddings, built-in web search or a complete multi-tool agent.

Using the OpenAI Chat Model node

n8n documents a Use Responses API toggle. Leave it off for this Chat Completions setup. Choose the custom provider base URL and model where your installed node exposes those settings. If that node version does not expose them, use the explicit HTTP Request starter instead of sending Standard Compute credentials to the default OpenAI host.

Separate the workflow bill from the model bill

n8n hosting or workspace charges and the API provider’s inference charges are separate. Standard Compute supplies a monthly model allowance; it does not include n8n hosting. A schedule that runs often, or a loop with retries, can use the allowance even when each individual response is short.

Make failures visible before enabling a schedule

Start with one item. Add a bounded retry policy for transient failures and an error path that you can inspect. Stop and check your account when its allowance is exhausted; retries cannot restore budget. Avoid sending the same record onward twice after retrying a model call. The starter is a connection template, not a tested production automation.

Run your own comparison

Use the same task, starting revision and acceptance check. Record every retry and failed attempt so the result reflects the work you actually paid for.

Sources and setup references