PRD Agent

Product Requirements Document management with AI agents

Draft PRD

AI-generated

Summary

Generate an initial Product Requirements Document (PRD) draft from a user-provided goal, optionally saving the output as markdown files to a specified directory.

Inputs

  • Goal (required) — The user goal or feature description that drives the generation of the PRD draft.
  • Project Name (required) — The name of the project to associate with the PRD.
  • PRD Output Directory — Directory path where the PRD markdown files will be saved; leave empty to skip saving files.
  • Working Directory — Directory where the Claude CLI command will be executed; useful for environment context.
  • Additional Options — Collection of additional parameters including Max Turns (maximum number of agentic turns, default 5), Verbose (enable detailed logging), and Model Alias (select the Claude model variant to use: Sonnet, Opus, or Haiku).

Output shape

a JSON object containing the operation, project name, optionally the saved markdown file path, raw and parsed AI response, session ID, usage statistics, and optionally detailed message logs if verbose is enabled

If a PRD Output Directory is specified, the generated PRD draft will be saved as markdown files there. The response includes both raw and parsed string output from the Claude AI model, along with usage metadata. Errors during markdown saving are caught and logged separately but do not prevent returning the AI response.

Examples

Example 1: Draft a PRD for a new feature

Set 'Goal' to describe the user feature, name the project, and optionally specify output and working directories and model options. The node returns the AI-generated PRD draft and saves it if configured.

Discussion