Resources · n8n integration

n8n Moonshot Kimi nodeConfigure Moonshot Kimi in n8n.

The n8n Moonshot Kimi node calls Kimi models straight from a workflow. It ships with the n8n AI package and exposes 2 operations, one for messaging a model and one for image analysis, each with its own parameters. Handy if n8n already runs your processes and you want Kimi output inside them.

Try the Moonshot Kimi node on n8nSee the 2 operations

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Moonshot Kimi node actually do?

It is an action node: on every item that reaches it, it sends a request to a Moonshot Kimi model and writes the answer back into the workflow data. Two resources are available, Text and Image, for a total of 2 operations. There is no sub-node port on it, so nothing gets plugged underneath it the way a chat model plugs under an agent.

First scenario, a support inbox. A message lands, the node runs text.message with a System Message that fixes the tone, and the reply goes out through Gmail after a human check. The conversation is built in the Messages collection, one entry per turn, with Role set to user or assistant.

Second scenario, documents. Photos of receipts sit in a binary field on the item, meaning a file carried alongside the JSON data. The image.analyze operation reads that field, answers the question you wrote in Text Input, and the extracted text drops into Google Sheets.

Third scenario, research notes. Web Search turned on lets the model look things up before answering, and the result is posted to Slack. Keep in mind that Web Search and Thinking Mode cannot both be on.

Most of the tuning happens in one place, the Options collection of text.message. System Message sets the behaviour once for the whole run. Frequency Penalty and Presence Penalty, both at 0 by default, push the model away from repeating itself or from circling the same topic. Include Merged Response, off by default, adds a single string merging all text parts of the answer, which saves a mapping step when the next node wants one field.

When to pick something else: this node answers a prompt, it does not decide which tool to call next. That job belongs to the AI Agent node, which takes a chat model as a sub-node, a node wired to one of its ports. Other providers have their own nodes too, such as OpenAI or Anthropic.

Known limits: the node is at version 1 and covers 2 operations, nothing else. No fine-tuning, no file management, no embeddings here. Moonshot bills its own API calls under its own terms, n8n adds nothing on top, and a request that fails on the provider side fails the execution.

Connect

What do you need before the node runs?

  1. 01

    Create the Moonshot credential

    Open the Credentials menu in n8n and add the Moonshot credential with your API key, the string the provider hands you for programmatic access. A credential is stored once and reused by every workflow on the instance, so a key rotation happens in one place and all your Kimi nodes follow.

  2. 02

    Or switch to Gateway credits on n8n Cloud

    On n8n Cloud, the credential field of the node offers Use Gateway credits. Pick it and the node runs without a Moonshot account and without your own API key. On a self-hosted instance, that option is not there, so the credential from step 1 is the way in.

  3. 03

    Select the model and run the node once

    The Model field is a resource locator: it queries the provider with your credential and lists the models it can see. If the list stays empty, the credential is the suspect, not the node. Run the node once on a single item to confirm the answer shape before you wire the rest of the workflow behind it.

Actions

Which operation covers your case?

The Moonshot Kimi node exposes 2 operations across 2 resources. For each one: the node as you configure it in n8n, the required fields, and our field notes.

Resources × operations matrix
ResourceAnalyze ImageMessage a Model
Image✓
Text✓

Image

1 operation
01

Analyze image

image.analyze

What you see in n8n

Notes & use cases

Reads one or several images carried on the item and answers a question about them in plain text. The images come from a binary field, the file attached to the workflow item, not from a URL.

Key parameters

  • Model (modelId): required, and it has to be a model that can read images.
  • Text Input: the question asked about the picture, default What's in this image?.
  • Input Data Field Name(s): the binary field holding the image, default data, several names separated by commas.
  • Maximum Number of Tokens: in Options, default 1024; lower it and the description gets shorter.
Use cases
reading a photographed delivery note, or sorting screenshots by what they show.

Text

1 operation
02

Message a model

text.message

What you see in n8n

Notes & use cases

Builds a conversation and returns what the model replies. Attachments are optional, and the generation settings sit in the Options collection.

Key parameters

  • Messages: one entry per turn, each with a Prompt and a Role, user to ask, assistant to set a tone.
  • Add Attachments: off by default; turn it on and Attachment Input Data Field Name(s) names the binary field to send.
  • Response Format: text for prose, json_object when the next node parses the answer.
  • Max Tool Calls Iterations: default 15, and 0 removes the ceiling.
Use cases
drafting a reply, or turning a free-text form entry into clean fields.
Need help

Need help automating Moonshot Kimi with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the n8n Moonshot Kimi node free to use?
Yes on the n8n side. The node belongs to the n8n AI package that ships with the product, so there is nothing to install and no extra charge, on n8n Cloud as well as on a self-hosted instance under the Community Edition and its Sustainable Use licence. A workflow behaves the same in both places. What is not free is the model itself: Moonshot bills the API calls you make under its own terms, exactly as it would outside n8n. The one exception is n8n Cloud, where Gateway credits let the node run without a Moonshot account of your own.
02What do you need for the node to work?
A credential and a model. The credential is created once from the Credentials menu with your Moonshot API key, then reused by every workflow on the instance. On n8n Cloud you can skip it and choose Use Gateway credits in the node's credential field instead. After that, the Model field is required on both operations, and it is a resource locator, so it pulls the list of available models through your credential. A required field left empty stops the node before it calls anything, which is usually the quickest explanation for an execution that fails right away.
03What are the limits of this node?
It covers 2 operations and nothing more: analyzing an image, and messaging a model. Version 1 is the maximum in the catalogue, and an older workflow may show fewer options. Two settings exclude each other, Thinking Mode and Web Search, so a run uses one or the other. Output length is capped by Maximum Number of Tokens, default 1024, and a long answer gets cut rather than continued. The node also has no sub-node port, so nothing attaches underneath it: it runs on each incoming item like any other action node.
04Which of the 2 operations should you start with?
Start with Message a model. It is the operation that covers most cases, it needs only a model and one message, and its Options collection is where you tune behaviour: System Message for instructions, Output Randomness (Temperature) at 0.7 by default for how varied the wording gets, Output Randomness (Top P) at 1 for how wide the sampling is. Move to Analyze image only when the content you care about is inside a picture. If your real need is a model that picks its own tools across several steps, the agent node is the better shape, not this one.
05n8n or Make for this kind of AI step?
It depends on where your data has to live. n8n runs self-hosted with Docker or npm, or on n8n Cloud, and the workflow is identical in both, so prompts and model answers can stay on infrastructure you control. Make is hosted only, with no self-hosting option, and is billed per operation, which is easy to forecast when volume is steady and less so when a loop fans out. Both give you a visual canvas. Pick on hosting, on data control, on the cost model, and on which editor your team reads faster.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations