Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mulerun.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

​
MuleRun-Ready n8n Examples

If you want to find some MuleRun-ready n8n workflows, check MuleRun n8n Examples

​
Supported n8n Versions

MuleRun currently supports the following n8n versions: 2.1.4, 1.119.2, 1.109.1, and 1.103.0. We recommend creators use the latest supported version within this range to build agents and workflows.

​
Currently Supported Node List

​
Input Nodes

  • n8n Form Trigger and Chat Trigger are supported. The n8n Form supports element type such as Dropdown, Checkboxes, Date, Text, File, etc. Chat Trigger now supports both text input and file uploads.
image.png
  • Please enter valid Accepted File Types when choose File Element Type in n8n Form Trigger.
image.png Note: the default size limit is 30MB for images and 50MB for all other file types. To set custom upload constraints, creators can add maxFileSizeMB (max ~200MB) and max_count (max ~50) to fields which fieldtype=file in n8n Form Nodes. MuleRun uses these for static, pre-execution validation of uploaded files. image.png
  • Please connect Memory Nodes such as n8n Simple Memory if you want your agent to maintain context across multiple turns when using Chat Trigger.
image.png
  • Other Trigger nodes, including Manual Trigger webhook, etc., are not supported.
  • The field names defined in the n8n Form will appear in the input section before the agent is executed. To ensure clarity, use meaningful and descriptive field names, or add hints to guide user input. Field names should follow either camelCase or snake_case conventions, and MUST NOT contain spaces.

​
Output Nodes

Creators can configure the output nodes and content freely. If a field name in any node matches the following names, it will be output in a specific format:
Field NameTypeUsageExample
markdown_contentstringDisplay Markdown text"# Title\nContent"
html_contentstringDisplay full HTML page"<!DOCTYPE html>..."
attachment_urlsstring/array/objectAttachment links (auto-extract URL)["https://..."] or text/object containing URL
navigation_urlsstring/array/objectAutomatically opened webpages (auto-extract URL)["https://..."] or text/object containing URL
binary_contentbinaryFile output (supports multiple)binary data
For specific display examples of these fields, please refer to Appendix 1: Examples of n8n Workflow Node Output. Final Node The Final Node also follows the field name display specifications mentioned above. Additionally, if the Final Node contains binary files or URLs, they will be automatically parsed and displayed. Error Message If the workflow fails during execution, error messages will be output. It is recommended to use n8n’s Stop and Error Node. This node can directly terminate the workflow, regardless of whether there are other end paths or final nodes. Workflows terminated through this node will display a ❌, indicate execution failure, and will not charge the user. The error_message from this node will be output on the MuleRun platform and displayed to the user.

​
LLM (Large Language Model) Nodes

It is recommended that Creators use the MuleRun Creator API Key to invoke Large Language Models. MuleRun currently supports:
  • OpenAI: Chat Completion and Response interfaces
  • Anthropic: Messages interface
In the n8n AI Agent node, you can use any LLM Chat Model node that supports the above message formats.
If you need to use the LLM API Key provided by the MuleRun platform (without requiring the creator to provide their own LLM API Key), it is recommended to use the AI Agent in n8n and connect it to an OpenAI Chat Model, as shown in the figure below.
image.png If you need to use non-OpenAI models, you can also specify them in the Model ID. For specific lists of supported models, please refer to: image.png

​
Intermediate Nodes

  1. All nodes officially supported by n8n are supported.
  2. If community nodes need to be included, please communicate in advance.
  3. Sub-workflows are not supported.

​
Currently Supported Credentials

​
n8n Official Nodes

  • All Credential nodes officially supported by n8n are supported.

​
OAuth Configuration

  • If the workflow requires OAuth-type Credentials, it is recommended that the Creator register as a developer with the OAuth service provider, create an application, and provide the ClientID and ClientSecret for that application.
  • The MuleRun OAuth Redirect URL is: https://mulerun.com/api/platform/agents/oauth2-credentials/callback
Non-Standard Credential Configuration Methods
  • If an API Key is used in the workflow, it must be configured using n8n’s Credential functionality. Hardcoding the Key in the workflow is not allowed. The n8n configuration is shown in the figure below:
image.png
  • Even if the API provider uses a custom API Key and n8n does not have a node for that request, it can still be configured using the Credential method.
ScenarioUsing CredentialsNot Using Credentials
Key Expiration/UpdateModify the configuration to update the API Key without updating the workflow codeUpdate the workflow code, submit for review, which requires an approval cycle
Key ProvisionCan be provided by MuleRun, Creator, or User, with multiple configuration optionsCan only be provided by the Creator, who must ensure the Key’s availability
Key SecurityEncrypted storage, no leakageMay be leaked to third parties, leading to Key consumption

​
Debugging Agent

To address the issue of creators lacking effective runtime data when debugging agents, we now supports custom parameters configuration of agent runtime output data for creators. Creators can configure settings for two different scenarios: MY CONFIG and AGENT USER.
  • MY CONFIG: These settings apply when the Creator runs or tests the agent. You can customize output levels and debug information to help with development and troubleshooting.
  • AGENT USER: These settings apply when End Users run the agent after it is published. This allows you to control the experience and information visibility for your users.
Four configuration options are supported for both scenarios:
  • Node Input/Output Information Level
  • Content Length Limits
  • Display LLM Node Model Token Usage Information
  • Display LLM Node Model Output Content
image.png

​
Message Output Level

Three levels of message output during the workflow execution are currently supported:
LevelNode Input/OutputNode StatusError Message
DEBUG✅✅✅
INFO❌✅✅
ERROR❌❌✅
Regardless of the selected level, the output message from the Final End Node will always be delivered as the workflow’s final result.

​
String Truncate Length

Maximum number of characters to show for workflow input/output data. Content longer than the String Truncate Length value will be truncated for better readability.

​
Model Token Usage Tracking

Shows model token consumption after LLM node executed.

​
Show LLM Generation Message

It will show the chat messages of LLM generation after LLM node executed.

​
Human In The Loop

We now support workflows that allow users to provide additional input while it is running. We call it Human-In-The-Loop (HITL). Creators can build workflows that include an n8n Form node to accept user input mid-execution. Currently, the n8n Form is the only node we support for HITL functionality. image.png HITL Form now supports showing dynamic data in workflow. image.png

​
Appendix 1: Examples of n8n Workflow Node Output

​
1. Dialog Box Output

​
Markdown Text

  • The final node output must have a JSON key field named markdown_content. The dialog box will render the Markdown effect. The node requirements are shown in the figure below:
image.png
  • The effect is shown in the figure:
image.png

​
Images

  • Also output using Markdown. The JSON field is markdown_content, and the content is: ![Example Image](https://example.com/image.jpg)\
  • The effect is as follows:
image.png

​
Videos

  • Similar to images, also in Markdown format. The JSON field is markdown_content, and the content is: <video controls>\n <source src="{{ $json.url }}">\n</video>
image.png

​
2. HTML Rendering

  • The supported output format is HTML file rendering. The JSON field is html_content, as shown in the figure below:
image.png
  • The effect is shown in the figure:
image.png
  • HTML rendering on MuleRun no longer requires creators to provide additional information, such as domain/subdomain or SSL certificates. MuleRun will now automatically host a unique subdomain for each HTML render.

​
3. Files

  • Any binary file data from the final node of the workflow (i.e., nodes not connected to any other nodes) will be uploaded to mulerun.com/workspace/files at the location: agents/{agent-name}-{time}-sid-{session-id}/{filename}
    • Node type: Unlimited
    • Node output field name: Unlimited
    • Binary file type: Unlimited
    • Filenames can be duplicated. If more than one file has the same name, a number will be appended to the end of the filename.
    • Filename length must not exceed 200 characters. Special characters are allowed.
    • Some n8n nodes cannot pass binary data from the input to the node output. In such cases, use a Merge Node to pass binary data across nodes and synchronously wait for data aggregation.
image.png
  • All items in the red box will be uploaded:
image.png image.png image.png

​
4. Navigation URLs

  • Set navigation_urls to https://mulerun.com will show like this:
image.png

​
Appendix 2: Example of Non-Standard Credential Configuration

​
Incorrect Approach: Hardcoded Key

  • Example: PiAPI
image.png
  • Requires including X-API-KEY: {YOUR-API-KEY} in the HTTP Header. Many Creators create an n8n HTTP Request node as follows:
image.png
  • And configure it as follows in the “Send Headers” section, filling in X-API-KEY: {YOUR-API-KEY} in the header:
image.png
  • Note: This is the wrong approach ❌. It makes it difficult for the Mule platform to detect that the Creator is using this credential, thereby complicating maintenance.

​
Correct Approach: Using Credentials

  • Create a Header Auth - type Credential in the credentials section, as shown in the figure:
image.png
  • Enter the Key - Value pair for the Auth. Name it piapi auth.
image.png
  • In the HTTP Request node, select Authentication > Generic Credential Type, then choose the Header Auth type and select the newly configured piapi auth.
image.png