n8n-nodes-alltoken

One API key. Dozens of frontier AI models. Works in n8n today.

Why AllToken

AllToken is a unified AI gateway that exposes a single OpenAI-compatible API across dozens of frontier models — Claude, DeepSeek, Gemini, Qwen, GPT-4o, and more. Instead of managing separate API keys and integrations for each provider, you get one key and one endpoint.

This node brings that directly into n8n:

• Chat — 46 models: Claude 4.5/4.6/4.7, DeepSeek v3/v4, Gemini 3.x, Qwen 3.5/3.6, Kimi k2.5/k2.6, GPT-4o/5.3/5.4/5.5, o3/o4, GLM 4.7/5/5.1, MiniMax, MiMo
• Image — generate images with gpt-image-2
• Video — generate video clips with seedance-2.0 / seedance-1.5-pro (async, 30 s – 3 min)
• AI Agent / Chain — the included Chat Model sub-node slots directly into n8n's built-in AI nodes; tool-calling works out of the box
• Web grounding — DeepSeek and Qwen models support an enable_search toggle for real-time web search

Install

1. Open n8n → Settings → Community Nodes
2. Click Install
3. Enter the package name: n8n-nodes-alltoken
4. Confirm

The two nodes (AllToken and AllToken Chat Model) appear in the node picker after installation.

(Screenshots in a later release.)

Get an API Key

Sign up at https://alltoken.ai, navigate to the API key page, and copy your key. Keep it — you'll paste it into the credential next.

Configure the Credential

1. In any workflow, click Add Credential → search AllToken API
2. Paste your API key
3. Click Save

n8n automatically tests the credential by calling /models. A green checkmark means you're good to go.

The Nodes

| Node | Type | Use when… | |---|---|---| | AllToken | Action node | You want to call Chat / Image / Video directly in a workflow | | AllToken Chat Model | Sub-node | You want to plug a model into an AI Agent or AI Chain |

AllToken — Chat

Select Resource → Chat, pick a model, write your prompt.

Minimal workflow: Manual Trigger → AllToken (Chat, model: deepseek-v4, prompt: "Summarize this text") → Set

Output shape:

{
  "content": "Here is the summary…",
  "model": "deepseek-v4",
  "usage": {
    "prompt_tokens": 42,
    "completion_tokens": 118,
    "total_tokens": 160
  }
}

Options available: system prompt, temperature, max tokens, enable_search (DeepSeek + Qwen only).

AllToken — Image

Select Resource → Image. Provide a text prompt; the node returns a binary item you can pipe directly into a Write Binary File or Send Email node.

Minimal workflow: Manual Trigger → AllToken (Image, model: gpt-image-2, prompt: "a futuristic city at dawn") → Write Binary File

Output shape:

{
  "id": "igen_...",
  "model": "gpt-image-2",
  "prompt": "a futuristic city at dawn",
  "status": "completed",
  "size": "1024x1024",
  "usage": { "input_tokens": 42, "output_tokens": 1 },
  "url": "https://...r2.dev/.../0.png"
}

url is a 30-day signed R2 URL — use it for quick preview or download. binary.data contains the actual PNG bytes as an n8n binary descriptor and is what downstream nodes (e.g. Write Binary File, Send Email) consume.

AllToken — Video (async)

Video generation is two steps because encoding takes time (30 seconds to 3 minutes).

Step 1 — Generate: Select Resource → Video, Operation → Generate. Provide your prompt and model. The node returns a task_id.

Step 2 — Poll / Get Result: Add a Wait node (e.g. 60 s), then a second AllToken node: Resource → Video, Operation → Get Result, paste the task_id from step 1.

Minimal workflow: Manual Trigger → AllToken (Video Generate) → Wait 60s → AllToken (Video Get Result) → Write Binary File

Get Result output shape:

{
  "status": "succeeded",
  "video_url": "https://…/video.mp4",
  "task_id": "abc-123"
}

Poll again if status is still processing.

AllToken Chat Model (sub-node)

This node does not run on its own. Drop it into the Chat Model slot of an AI Agent or Basic LLM Chain node.

How to use:

1. Add an AI Agent node to your workflow
2. Click the Chat Model input connector
3. Add AllToken Chat Model
4. Select a model, optionally set temperature / max tokens
5. The agent handles tool-calling automatically via langchain

This is the easiest way to swap between dozens of models inside a single agent workflow without changing anything else.

Common Errors

| HTTP code | Meaning | Fix | |---|---|---| | 401 | Invalid API key | Check the credential — re-paste your key | | 402 | 余额不足 (insufficient balance) | Top up your AllToken account at alltoken.ai | | 410 | Image result expired | Image results are one-shot; regenerate the image | | 429 | Rate limit exceeded | Wait a moment and retry; reduce request frequency |

Roadmap

• More image models as AllToken expands its image catalogue
• Additional video models
• n8n verified node status

Development

For contributors who want to build or test locally:

npm install
npm run build
npm test

To run a local n8n instance with the node loaded:

docker compose up -d

The docker-compose.yml mounts dist/ as a custom extension directory so n8n picks up your build automatically.

License

MIT — see LICENSE.