Frameworks & Platforms
n8n
Use TokenLab chat models as the language model inside n8n AI Agent workflows
Overview
n8n runs AI Agent workflows through the LangChain-style node set: an AI Agent node takes a Chat Model sub-node. The community node package n8n-nodes-tokenlab provides that chat model for TokenLab, so one credential and one Base URL cover every chat model in the catalog.
The package contains two pieces:
- Credential TokenLab API (
tokenlabApi): a Base URL and an API key. - Sub-node TokenLab Chat Model (
tokenLabChatModel): outputsai_languageModel, and loads the model list fromGET /v1/models?category=chat.
Install
Community nodes have to be enabled on the instance first: Settings → Community Nodes, turn on community nodes (N8N_COMMUNITY_PACKAGES_ENABLED=true for self-hosted Docker).
n8n-nodes-tokenlab is an unverified community node, so it installs on self-hosted n8n only: n8n Cloud only lists verified community nodes. On a self-hosted instance, open Settings → Community Nodes → Install and enter n8n-nodes-tokenlab as the package name.
For a self-hosted instance where the package is baked into the image instead, add n8n-nodes-tokenlab to the instance dependencies and restart n8n; installing it with npm inside the n8n container alone does not register the node.
Create the credential
- Create an API key in the TokenLab dashboard.
- In n8n create a TokenLab API credential:
- Base URL:
https://api.tokenlab.sh(change it only if you call through your own proxy). - API Key: the key from step 1.
- Base URL:
- Save. n8n sends the key as
Authorization: Bearer <API key>and tests it by listing chat models.
Use it in a workflow
Add an AI Agent node
Add AI Agent to the canvas.
Attach the model
Add TokenLab Chat Model and connect it to the Agent with the ai_languageModel connection type.
Pick a model
Choose a model from the dropdown (it is loaded from your TokenLab catalog) or type a model ID directly, for example claude-sonnet-5-5, claude-haiku-4-5 or qwen3.8-max.
Run
Execute the workflow. Every run is billed to your TokenLab account at the same per-token price as the API.
A minimal workflow export with the two nodes wired together:
{
"nodes": [
{
"parameters": {},
"id": "agent-1",
"name": "AI Agent",
"type": "@n8n/n8n-nodes-langchain.agent",
"typeVersion": 1.7,
"position": [0, 0]
},
{
"parameters": { "model": "claude-sonnet-5-5", "temperature": 0.7, "maxTokens": -1 },
"id": "model-1",
"name": "TokenLab Chat Model",
"type": "n8n-nodes-tokenlab.tokenLabChatModel",
"typeVersion": 1,
"position": [-240, 0],
"credentials": { "tokenlabApi": { "id": "REPLACE_WITH_CREDENTIAL_ID", "name": "TokenLab API" } }
}
],
"connections": {
"TokenLab Chat Model": {
"ai_languageModel": [[{ "node": "AI Agent", "type": "ai_languageModel", "index": 0 }]]
}
}
}Choose a model
The dropdown lists the chat models your account can call. Copy the exact ID from the Models page when you want to pin a specific model ID. Non-chat modalities (images, embeddings, video) are not covered by this package; call them with the HTTP Request node against https://api.tokenlab.sh/v1.
Troubleshooting
| Symptom | Fix |
|---|---|
TokenLab Chat Model does not appear in the node list | Community nodes are disabled on the instance, or the package is not installed; reinstall n8n-nodes-tokenlab and restart n8n |
| Credential test fails with 401 | The key was deleted or copied with a trailing space; create a new key in the dashboard |
| Model dropdown is empty | The key has no model access on that account; check the account's catalog on the Models page |
| 429 from a workflow run | The account hit its rate limit; see Rate limits |