Skip to main content
n8n is a leading workflow automation platform with built-in AI capabilities — AI Agent nodes, LLM chains, and RAG components. Its OpenAI credential supports a custom Base URL, so every OpenAI-flavored AI node can run on the ScitiX Model Inference API (same steps on n8n Cloud and self-hosted). See the Models page for available models (e.g. glm-5.2).

​
1. Create the credential

Create an API key on the ScitiX API Keys page. Then in n8n, open the Credentials tab, click Create credential, and pick OpenAI in the app selector. Fill in: When you save, n8n tests the credential with a GET /models call against the Base URL — Connection tested successfully means ScitiX is reachable and the key works. n8n — ScitiX (OpenAI) credential, connection tested successfully

​
2. Use it in a workflow

  1. Create a workflow with a Chat Trigger (“On chat message”) and an AI Agent node.
  2. Open the AI Agent node, click + Chat Model, and pick OpenAI Chat Model.
  3. In the OpenAI Chat Model node, select the ScitiX credential (on n8n Cloud, choose My own credential first). The Model dropdown loads live from ScitiX’s /models endpoint — select glm-5.2. The AI Agent node runs as a Tools Agent since n8n 1.82.0, so the model must support native tool calling; glm-5.2 does. Check the platform’s tool-calling list before switching models.
  4. Leave Use Responses API off — that toggle targets OpenAI’s own API.
  5. Connect at least one tool sub-node to the AI Agent (a documented requirement — e.g. the Calculator tool).
The credential is shared by the OpenAI Chat Model, Embeddings OpenAI, and plain OpenAI nodes. For ScitiX, prefer the Chat Model sub-node: many plain-OpenAI-node operations (image, audio, files) call OpenAI-specific endpoints. For embeddings-based workflows (RAG), add an Embeddings OpenAI node on the same credential and pick an embedding model from the Models page.

​
Verify

Open the chat panel and send a message that exercises the tool, for example “what is 17 * 23?”. A streamed answer with a visible tool call and the result 391 confirms both the connection and tool calling. For a plain completion test without tools, use a Basic LLM Chain root node with the same OpenAI Chat Model sub-node. n8n — AI Agent using the Calculator tool via ScitiX, answering 17 × 23 = 391

​
FAQ

  • The Model dropdown is empty: The GET /models call failed. Re-check the Base URL (/v1 suffix, no trailing slash) and the key.
  • Which tool should I use for a quick test? Prefer the Calculator tool. If you use a Code tool instead, each parameter must be individually enabled for the AI to fill (click the ✨ on it, or you get No parameters are set up to be filled by AI), and the Code node runs on a separate task runner that times out at 60s if that runner isn’t enabled. Calculator avoids both.
Verified on self-hosted n8n 2.32.7.