Super Agent icon

Super Agent

Tool-calling AI agent that responds to user queries using connected language models and tools
Join our community

Overview

The "Super Agent" node is an AI-powered agent designed to handle complex tasks by dynamically routing prompts through connected language models and orchestrating multiple tools. It maintains conversation context using memory connections and can leverage specialized prompt routing for tailored responses. This node is particularly useful when you want to automate multi-step workflows that require reasoning, tool usage (e.g., calling APIs or functions), and contextual understanding over a conversation.

Common scenarios:

  • Automating customer support where the AI uses external tools to fetch data or perform actions.
  • Building intelligent assistants that maintain stateful conversations and invoke various services dynamically.
  • Complex decision-making workflows where the AI needs to call different tools based on user input.
  • Using prompt routing to delegate parts of the conversation to specialized sub-agents or prompts.

Practical example:
You could connect a chat language model, a memory module to keep track of conversation history, and several tools like a calendar API, weather service, or database query interface. The Super Agent will receive user input, decide which tools to call, execute them, and return a comprehensive response while maintaining context across interactions.

Properties

Name Meaning
MCP-Powered AI Agent with Dynamic Routing & Tool Orchestration Informational notice describing the node's capabilities: executing complex tasks using connected tools, maintaining conversation context with memory, and leveraging prompt routing for specialized responses.
Prompt How to provide the prompt text to the agent:
- Take From Previous Node Automatically: Use the input from the previous node as the prompt.
- Define Below: Manually specify the prompt text in the next field.
Text The manual prompt text to send to the agent (shown only if "Define Below" is selected). Example: "Hello, how can you help me?"
System Message Source Choose the source of the system message that defines the agent's role and behavior:
- Use Super Agent System Message: Use the system message defined in this node.
- Route via Super Prompts: Delegate to connected Super Prompts node.
System Message The system message content defining the agent’s role and behavior (shown only if "Use Super Agent System Message" is selected). Default is a helpful AI assistant description instructing tool usage and task completion.
Enable Output Parser Whether to use a connected Output Parser node to process the AI response before outputting it.

Output

The node outputs an array of items, each containing:

  • json object with:
    • All original input fields merged.
    • output: The final response from the AI agent after processing, including any tool calls and memory context.
    • chatInput: The prompt text sent to the AI (either auto-extracted or manually defined).

If the Output Parser is enabled and connected, the output field contains the parsed result instead of raw AI text.

The node does not output binary data.

Dependencies

  • Requires connection to at least one AI language model node that supports function calling or invocation.
  • Optionally connects to:
    • Memory nodes to maintain conversation context.
    • Tool nodes that expose callable functions for the agent.
    • Output Parser node to post-process AI responses.
  • Proper configuration of API keys or credentials for the connected AI models and tools is necessary.
  • If using external system message routing, a connected Super Prompts node must be configured.

Troubleshooting

  • No Language Model Connected Error:
    The node requires a connected AI language model. Ensure you have connected a compatible language model node.

  • Model Does Not Support Tools / Function Calling:
    Some models do not support function calling or tool integration. Common error messages indicate unsupported features or bad requests.
    Resolution: Use supported models such as GPT-4, GPT-3.5-turbo, or equivalent models from other providers that support function calling.

  • Tool Execution Errors:
    If tools are not formatted correctly or missing required methods, tool execution may fail. Check that all connected tools implement the expected interface (invoke method).

  • Memory Save/Load Failures:
    Memory nodes might not implement known interfaces for loading or saving chat history. Verify that memory nodes conform to expected methods like getChatMessages, saveContext, or addMessage.

  • Output Parsing Issues:
    If output parsing is enabled but no parser is connected, a warning is logged and raw output is returned. If parsing fails, errors are logged and raw output is used.

  • Prompt Extraction Problems:
    When using automatic prompt extraction, ensure the input data contains recognizable fields like chatInput, text, or message. Otherwise, the prompt may be empty or malformed.

Links and References

Discussion