Skip to content

n8n: using the VIDAI Control Plane as the backend

n8n's OpenAI credential has a Base URL field. Point it at the control plane and every OpenAI node (chat, embeddings, image generation, assistant) in every workflow uses your VIDAI endpoint.

TL;DR

  1. Credentials → New → OpenAI account.
  2. Fill in:
  3. API Key: your VIDAI API key
  4. Base URL: https://your-vidai-server.example.com/v1
  5. Save.
  6. In any OpenAI node, pick this credential. The Model dropdown will show models the control plane returns (or you can pick "custom" and type a registered model name).

Prerequisites

  • n8n installed (cloud, Docker, or bare).
  • Control plane base URL and an API key from API Keys.
  • A model registered on the Models page.

Which nodes route through the control plane

Any node that uses your VIDAI-configured OpenAI credential:

  • OpenAI node (Message, Assistant, Image, Audio, Embedding operations)
  • AI Agent node when its Chat Model is set to the OpenAI Chat Model sub-node using your credential
  • AI Chain nodes (LangChain-based) that use the OpenAI Chat Model sub-node with your credential

n8n's other LLM nodes (Anthropic, Google, HuggingFace) each have their own credentials — configure the Anthropic credential separately if you use it (see below).

Anthropic credential

n8n's Anthropic credential accepts a Base URL too. Note the Anthropic base URL is the root (no /v1):

  • API Key: your VIDAI API key
  • Base URL: https://your-vidai-server.example.com

Attribution

One VIDAI API key per n8n workflow (or per team) — the credential is the attribution unit. If several workflows share a credential, they share a Chargeback line. Split credentials to split lines on the Chargeback tab.

AI Agent nodes

n8n's AI Agent node handles multi-turn agent reasoning. When its Chat Model is set to your VIDAI credential, every reasoning step and every tool call flows through the control plane. Common tools (HTTP Request, code, calculator) are executed inside n8n — the LLM decides which to call, and each decision is a control-plane request.

Verify it works

Build a two-node workflow:

  • Manual Trigger → OpenAI (Message a model) with your VIDAI credential, model gpt-4o-mini, prompt reply with exactly: ok.

Run once. Confirm the response. A row appears on Request Logs.

If something's off

Raise an issue at github.com/vidaiUK/vidai-quickstart/issues with the credential setup (masking the key) and the failing node execution log. We'll get it sorted.

Where to go next