Claude Code: set up research and content agents

This tutorial explains how the presenter builds an AI agent workflow in Claude Code through the Claude Desktop app. The setup uses a project folder and plain English instructions rather than application code. It covers files in a desktop workspace; it does not demonstrate running a model on your own hardware.

The central file is CLAUDE.md. The presenter recommends recording project context, audience, writing preferences and working rules there, alongside folders for workflows, reference material and completed output. Each workflow gets a separate Markdown document with its goal, steps, expected output and instructions for handling problems. The suggested process requires clarification and a written plan for approval before execution.

Two examples put that structure to use. A research workflow asks about scope and audience, proposes a plan, then saves a report. The example targets roughly 1,500 words about AI agents in 2026. The presenter describes a repurposing workflow intended to turn one script into three 60-second scripts and a social media pack, with four Markdown files converted to PDFs.

The presenter advises refining one workflow before connecting several together. Claims about speed, persistent context and output quality are the speaker's descriptions, rather than independently verified guarantees. Separate sections promote sponsor GoHighLevel and the speaker's AI Master platform; neither is established as a prerequisite.