n8n

The OpenAI credential's Base URL field — https://api.kunavo.com/v1, keeping the /v1 — puts every OpenAI Chat Model in an n8n workflow on Kunavo; the model node itself has no endpoint field.

Credentials  →  Create credential  →  OpenAI
  API Key                      sk-kn-...
  Organization ID (optional)   leave empty
  Base URL                     https://api.kunavo.com/v1     <- keep the /v1

Workflow  →  AI Agent or Basic LLM Chain  →  Chat Model: OpenAI Chat Model
  Credential to connect with   the OpenAI credential above
  Model                        ID mode:  claude-sonnet-5
  Use Responses API            on   → POST /v1/responses
                               off  → POST /v1/chat/completions

Step by step

1. Create a key at /app/keys and copy it — it is shown once.
2. In n8n, create a credential of type OpenAI. Put the key in API Key, leave Organization ID (optional) empty, and replace the Base URL default https://api.openai.com/v1 with https://api.kunavo.com/v1. Save.
3. Add an AI Agent or Basic LLM Chain node, and attach an OpenAI Chat Model sub-node using that credential. Switch the Model field from From List to ID and type the id the way GET /v1/models lists it, for example claude-sonnet-5 — the list works too, but a typed id keeps the workflow readable.
4. Decide on Use Responses API: leave it on unless a tool in your chain expects Chat Completions, or you want the n8n execution log to show a chat-completions request.
5. Run the workflow once with a one-line prompt before wiring it to a trigger. A 401 means the key; a 404 whose message begins with <!DOCTYPE html> means the Base URL lost its /v1.

Checked against n8n's OpenAI credential source at tag n8n@2.41.4 on October 1, 2026. Third-party settings move; if a field name here no longer matches what you see, that page is the authority, not this one.

Verify before you debug the client

One request settles whether a failure is the endpoint, the key, or the configuration file. If this returns JSON, the same base URL and key work in n8n.

# Settles whether a failure is the endpoint, the key, or the client.
curl -sS https://api.kunavo.com/v1/models \
  -H "Authorization: Bearer sk-kn-..."

Which model id to put in the field

Every text model is reachable as a model id — the live list is GET /v1/models, and the catalog with prices is on the models page. Rates are USD per 1M tokens, input / output.

Why the Base URL is on the credential

Older tutorials set the endpoint inside the model node. In the released source that in-node Base URL option is hidden from node version 1.1 onward, so a node you add today has no such field and the credential's Base URL is the one that counts. n8n's own OpenAI Chat Model documentation and its credential page do not describe the field; the source does, with the description “Override the default base URL for the API”. The HTTP Request node is a different route entirely — it works, but you would be hand-building the request an AI Agent node builds for you.

Responses API on or off

• On (the default on node 1.3) — requests go to /v1/responses. This is the only mode that shows the node's Built-in Tools — Web Search, File Search and Code Interpreter. Those are OpenAI-hosted tools; nobody has tested them through Kunavo, so do not build a workflow that depends on them without trying it first.
• Off — requests go to /v1/chat/completions, the most widely supported shape, and the one to fall back to if a tool call misbehaves under the other.
• Tools you attach to an AI Agent are sent to the model as function definitions. That round trip was not part of this check, so run one tool call on a test workflow before relying on it.

n8n and OpenRouter

n8n ships a separate OpenRouter Chat Model node with its own OpenRouter credential. That credential has an API Key field and a Base URL that is hidden and fixed at https://openrouter.ai/api/v1, and its test calls OpenRouter's own /key route — so the OpenRouter node can only ever talk to OpenRouter. If OpenRouter is what you want, use that node with an OpenRouter key; nothing on this page is needed.

Any other OpenAI-compatible endpoint, Kunavo included, goes through the OpenAI Chat Model and the OpenAI credential's Base URL as above. Choose between them on what actually differs — which models you need, how you want to pay, whether you want one balance across n8n and your other tools — not on the node. Kunavo's side of that comparison is in Kunavo vs OpenRouter.

Keeping an unattended workflow's bill bounded

• The node's Max Retries defaults to 2 and Timeout to 60000 ms. A request that times out is retried, and a retry is a new billed request.
• Set Maximum Number of Tokens on nodes that run per item — a loop over 1,000 rows multiplies whatever one call costs.
• Use a separate Kunavo key for each production workflow so the usage page shows which one spent what, and revoke one without touching the others.

What the errors look like

• “401 Missing or invalid API key” — the Base URL is right and the key is not. Reproduced on 2.41.4.
• “404 <!DOCTYPE html>…”, filed by LangChain under MODEL_NOT_FOUND — misleading: the model is fine, the Base URL is missing /v1 and the request landed on the website. Reproduced on 2.41.4.
• A model-not-available message in JSON — the model ID does not match GET /v1/models exactly.