Pushbuild

How to use Kimi and other OpenAI-compatible models in n8n

Many model providers, Kimi (Moonshot), DeepSeek, OpenRouter and local servers such as Ollama among them, accept the same API format as OpenAI. n8n does not need a separate node for each one. You point the OpenAI credential at the provider's address and type the model name.

At a glance

  1. Set up the credentialCreate a new OpenAI credential in n8n.
  2. Use it in a workflowIn the OpenAI Chat Model node (for Agent node and chains) or the OpenAI node, select the new credential.
  3. Common problems401 or 403: the key belongs to a different provider than the Base URL.

Set up the credential

  • Create a new OpenAI credential in n8n. Paste the provider's API key, not an OpenAI key.
  • Change Base URL to the provider's OpenAI-compatible address, as given in their API docs. It usually ends in /v1.
  • Save. The connection test may fail if the provider does not list models the way OpenAI does; the credential can still work.

Use it in a workflow

  • In the OpenAI Chat Model node (for Agent node and chains) or the OpenAI node, select the new credential.
  • If the model list is empty or shows Openlanguage models, switch the model field to an expression or ID and type the provider's model name exactly as their docs give it.
  • Run one test message first, then check the output for the provider's own model name.

Common problems

  • 401 or 403: the key belongs to a different provider than the Base URL.
  • 404: the Base URL is missing /v1, or the model name is wrong.
  • Tool calls not working: not every compatible model supports function calling. Check the provider's docs before using it with the Agent node.
  • A local model in Docker: use the host's address, not localhost, because localhost inside the n8n container is the container itself.

Updated 2026-11-01. General guidance, not advice for your situation; we scope each project in writing before any work starts.

Questions people ask first

Do I need a community node?

Not for providers that offer an OpenAI-compatible endpoint. The built-in OpenAI nodes work with a changed Base URL.

Can I use several providers in one workflow?

Yes. Create one credential per provider and choose the credential in each node.

Will image and audio features work?

Only if the provider supports those endpoints. Chat is the most widely supported.

Tell us what you need built.

Tell us the process and the tools involved, or call. We reply with questions or a written scope.

hello@pushbuild.com

(617) 299-8887