Secure AI Agent

Secure AI Agent with Guardrails

Secure AI Agent

AI-generated

Summary

Send a user-defined or previous node prompt through a security guardrail that redacts sensitive information, processes the prompt securely, and restores redacted content in the response.

Inputs

  • Source for Prompt (required) — Choose whether to define the prompt text directly in the node ('Define Below') or automatically use the input from the previous node ('From Previous Node').
  • Prompt (User Message) — Enter the text prompt to process when 'Define Below' is selected as the source for prompt.
  • Options — Additional parameters including an optional system message to contextualize the AI response, and a toggle to require a specific output format.

Output shape

a list with a single JSON object containing an 'output' field

The output JSON contains 'output', a string with the secure AI response where sensitive data has been masked during processing and restored afterwards. The operation requires a connected Security Guardrail tool node to enable redact and restore functionality, else it will error.

Links

Discussion