n8n
Call Atlas Cloud LLM, image, and video models from n8n workflows using the community nodes.
The Atlas Cloud community nodes bring LLM chat and asynchronous media generation into n8n, so you can wire model calls into automations alongside your other integrations.
npm package: n8n-nodes-atlascloud
Installation
In n8n, go to Settings → Community Nodes → Install.
Enter the package name:
n8n-nodes-atlascloudConfirm the install and refresh n8n. The Atlas Cloud nodes appear in the node panel.
Community nodes are only installable on self-hosted n8n instances, and require the instance owner's permission.
Adding credentials
Create an API key at console/api-keys.
In n8n, create a new credential of type Atlas Cloud API.
Paste the key and save. Every Atlas Cloud node in your workflows can now reference this credential.
The nodes
Atlas Cloud Chat
Calls the OpenAI-compatible chat endpoint.
| Field | Purpose |
|---|---|
| System Prompt | Sets model behavior |
| User Prompt | The request |
| Temperature | Sampling randomness |
| Max Tokens | Output length cap |
| Extra Messages | Additional conversation turns for multi-step context |
Atlas Cloud Task Submit
Submits an image or video generation job. The node discovers available categories and models, then renders the parameter form from the selected model's schema — so you get the right fields for that specific model rather than a generic form.
Two modes:
| Mode | Behavior |
|---|---|
| Submit Only | Returns the prediction ID immediately. Pair with Task Status later |
| Wait for Completion | Blocks until the job finishes and returns the result |
For video models, prefer Submit Only. Generation can take minutes, and holding an n8n execution open that long is fragile — a queue-and-poll design survives restarts and timeouts much better.
Atlas Cloud Task Status
Checks a job by prediction ID. Configurable polling interval, timeout, and retry count.
A typical asynchronous workflow: Task Submit (Submit Only) → store the ID → Wait → Task Status → branch on the result.
Example workflows
The repository includes importable graphs under workflows/:
| File | Pattern |
|---|---|
01-chatbot.json | LLM chat responder |
02-text-to-image.json | Synchronous image generation |
03-async-video.json | Submit, poll, and handle the finished video |
Notes
- The nodes are maintained in the open-source repository rather than shipped with n8n, so they lag the newest models by some margin. If a model you need is missing from the picker, call the HTTP Request node against the API directly.
- Generation is billed the same as any other API call. See Model Billing.
- For long-running video jobs, webhooks are more reliable than polling — point the webhook at an n8n Webhook trigger node.
Related
Last updated on