Connect n8n to an OpenAI-compatible API

Connect n8n to an OpenAI-compatible API using AELDER’s key and endpoint. This guide uses an OpenAI credential and the OpenAI Chat Model sub-node with a Basic LLM Chain. It gives you a small text workflow to check before building a larger automation or AI agent.

1. Prepare your key and model

Use an AELDER API key, the service Base URL shown above, and an exact model ID from your account’s model list. Your n8n administration API key is a different credential: it manages n8n and cannot authenticate model requests.

Review the model catalog and token prices before choosing. For DeepSeek in n8n, the DeepSeek V4.1 Flash page explains that catalog model; confirm its exact ID and availability in your account before using it.

To compare releases before selecting one, use the DeepSeek API model and price table. It keeps the 1017 entry distinct from the other catalog versions.

2. Create an OpenAI credential

In n8n, create a new OpenAI credential and give it a recognizable name such as AELDER. Enter these values before saving or testing it:

Keep the key in n8n’s credential storage. Do not paste it into a workflow prompt, Code node or exported JSON. On n8n Cloud, select your own credential for this connection rather than Gateway credits.

The credential test checks the models endpoint. A successful check does not prove that a selected model can generate a reply or use tools. If it fails, confirm that the credential’s Base URL points to AELDER before changing anything else.

3. Build a small text workflow

Add a Manual Trigger and connect it to a Basic LLM Chain. In the chain, set the prompt source to Define below and enter Reply with OK. in Prompt (User Message). Leave Require Specific Output Format off for this first check.

Use the chain’s Model connection to attach an OpenAI Chat Model. The model is a sub-node: connect it to the chain’s model port, not as a separate step on the main workflow line.

Manual Trigger → Basic LLM Chain
                  Model: OpenAI Chat Model

4. Configure the chat model

Select the AELDER credential in OpenAI Chat Model. Switch the Model selector to ID and enter the exact account model ID, particularly if a custom model is missing from the dropdown.

• Keep Use Responses API off where that toggle is shown. This recipe uses Chat Completions.
• Under Options, set Max Retries to 0 for the initial check.
• If an older node version exposes a node-level Base URL override, leave it unset or use the same service Base URL. A conflicting override can send generation to a different endpoint from its credential test.
• Start with text output and default capability settings. Add reasoning, structured output or tools only after checking support for the selected model.

When ready, execute the workflow once. This sends a billable generation request. Inspect the chain’s output, then confirm the model and token usage in your AELDER dashboard. Check the failure details before running it repeatedly.

Using the model with an AI Agent

An n8n AI Agent also needs a connected chat-model sub-node. Use a model that supports the agent’s tool-calling requirements, and configure the tools you intend it to call. A Basic LLM Chain text reply does not establish tool compatibility.

Test one small tool interaction before adding a larger workflow. Conversation history, tool results and repeated model calls can increase token usage. Keep the first check separate from workflows that send messages or modify external records.

n8n workflow and model API costs

Your n8n plan or hosting and AELDER’s model API usage are separate costs. A workflow execution can make more than one model request, especially with agents, loops or retries.

Compare input, output and cached-input rates for the selected model. Use the API cost calculator for an estimate, then check actual usage for your workflow. A lower token rate alone does not determine the total cost of an AI agent run.

Troubleshoot the connection

Reference: n8n’s OpenAI Chat Model documentation. Node and credential settings were reviewed; live AELDER compatibility has not been verified.