AI Agent (Raevon)

Universal AI Agent - works with ALL n8n chat models (native & community), correct tool calling, intermediate webhook

AI Agent (Raevon)

AI-generated

Summary

Execute an AI agent that uses a connected chat model, optional memory, tools, and output parser to generate intelligent responses based on a user prompt, managing tool invocation and conversation memory with support for intermediate webhook updates and configurable system instructions.

Inputs

  • promptType (required) — Select 'Connected Chat Trigger Node' to automatically use the 'chatInput' field from a connected chat trigger node, or 'Define below' to enter a static or dynamic prompt text manually.
  • text (required) — The user message prompt text to send to the AI agent. When promptType is 'auto', this defaults to the input's 'chatInput' field; when 'define', use this field to provide the prompt.
  • options.systemMessage — Optional system message that defines the AI agent's behavior and instructions, enabling customization of its assistant persona.
  • options.maxSteps — Maximum number of tool call steps the AI agent is allowed to perform before stopping. Defaults to 5, range 1-20.
  • options.intermediateWebhookUrl — If set, the node will send JSON POST requests to this URL with partial replies and tool calls as the AI agent processes the input, useful for real-time monitoring.
  • options.useOutputParser — If enabled, the node expects a connected output parser node to format the final AI output response. The agent uses this parser to ensure structured final output.
  • options.verboseLogs — Enables detailed console logging for debugging the AI agent's internal operations and tool interactions.

Output shape

a single JSON object per input item containing 'output' (the final AI response, parsed if using an output parser), 'steps' (an array of intermediate step details including tool calls), and 'totalSteps' (the count of steps executed).

Supports multiple input items processed sequentially with error handling. Throws errors if no language model is connected or output parser is enabled without a connected parser. Conversation memory is loaded and saved if an AiMemory node is connected, preserving context across runs. Intermediate webhook and verbose logs are optional features.

Links

Discussion