Module 1 – You’re Not Chatting, You’re Collaborating

The mindset shift is simple: stop thinking of AI as a search engine and start thinking of it as a configurable collaborator.

In a normal chat app, you type, you get a generic answer, done. With an AI agent, you get to define who the AI is, what it knows about you, and how it should respond. You provide the context – and the more relevant context you give, the better the output.

Professional analogy: It’s the difference between hiring a generic freelancer vs briefing a team member who knows your company, your industry, your clients, and your standards. The freelancer needs everything explained from scratch every time. The team member already has context and gets better over time.

Home analogy: It’s like asking a stranger for parenting advice vs talking to a friend who knows your family, your values, and your kids’ personalities.

You decide what context matters. At work, that might be your role, your industry, your processes, your compliance requirements. At home, it might be your family’s values and routines.

Module 2 – Your Agent’s Brain: What .md Files Are and Why They Matter

AI context is structured in .md (Markdown) files – simple text files the AI reads to understand its job. They’re just text with simple formatting. Nothing scary. These files are the “training manual” you write for your AI agent.

A key advantage: these files live on your computer. They’re plain text – not locked into any platform. If you build your agent on Claude today and want to switch to a different tool tomorrow, you take your files with you. No vendor lock-in, no export headaches. This portability is one of the most important reasons to structure your agent this way.

Module 3 – One Agent, One Job (Build Your First Agent)

An agent is an expert. Just like at work each person has a role with a specific expertise, your AI agent works the same way. You wouldn’t ask your accountant about marketing messaging, and you wouldn’t ask the marketing strategist to create social media visuals. Start with one agent, one clear purpose, well-structured files.

Pick something concrete:

• Work examples: a “client proposal drafter,” a “weekly status report writer,” a “meeting prep assistant,” a “research summarizer”
• Home examples: a “homework support advisor,” a “family meal planner,” a “screen time discussion helper”

Walk through creating the basic file structure: write a simple IDENTITY.md (who is this agent?), a simple SOUL.md (what are its values and boundaries?), and a USER.md (who are you and what does the agent need to know?). Key principle: don’t duplicate content across files – each file has its own job, just like the agent has its job.

Module 4 – Making It Smarter Over Time

The .md files aren’t write-once-and-forget. This is an ongoing relationship. The loop: use the agent on a real task, notice where it gets things wrong, refine the files, try again. This isn’t machine learning “training” – it’s more like onboarding a new team member.

Common refinements:

• The agent is too generic – add more specifics to IDENTITY.md
• The agent forgets important context – update MEMORY.md
• The agent doesn’t follow your preferred approach – clarify in AGENTS.md
• The agent crosses boundaries you care about – tighten SOUL.md

You don’t have to update the files manually. You can ask the agent to update its own skills and files based on what it just learned. For example: you refined a process together, or you finally nailed the tone of voice you want – ask the agent to save that into its own files. It’s like a trainee writing their own notes after a training session.

Module 5 – When Agents Talk to Your Tools (Connectors)

Connectors let your agent interact with other apps – read from them, write to them, act on them. The most common protocol right now is MCP (Model Context Protocol) – an open standard that’s becoming the default way agents talk to tools.

Your work doesn’t live in one place. Your projects are in Notion, your calendar is in Google Calendar, your emails are in Gmail, your designs are in Canva. Connectors let the agent work across those tools instead of you copying and pasting between them.

Key connectors:

• Notion – the most important one for getting started. It speaks the same language as your agents. Think of it as a structured input/output layer.
• Google Calendar / Gmail – scheduling, email drafting, meeting prep
• Design tools (Canva, Figma) – generating visual assets, creating presentations
• Project management (Linear, Asana) – creating issues, updating status, tracking work

Module 6 – The Notion Bridge: Your Shared Workspace with AI

Notion pages become the conversation surface between you and your agent. You can write a comment on a Notion page and the agent will act on it.

Example workflows:

• Marketing: Comment on a campaign brief: “draft three LinkedIn post variations” – agent creates them inline. Comment: “tone is too formal” – agent rewrites.
• Project management: Comment on a status page: “pull completed tasks and draft the client update” – agent compiles and drafts.
• Content: Comment on a blog draft: “add a section on compliance implications” – agent researches and adds.
• Home: Comment on a meal plan: “swap Thursday’s dinner for something vegetarian” – agent updates the plan.

Train your agent on your Notion structure. You have two approaches: (1) ask the agent to learn your existing Notion structure and save it to its SKILLS.md, or (2) ask the agent to recommend a structure for your use case and refine it together.

Module 7 – Multiple Agents, One Team (and One Household)

Once you’re comfortable with one agent, you might want more – each expert in something different. The key principle: each agent should be an expert in one thing.

You wouldn’t ask your accountant to also handle your marketing messaging. When you give one agent too many jobs, it gets confused about which hat to wear. Wisely separate and define what each agent does. Shared context goes in shared files; specialized context goes in each agent’s own files.

Module 8 – Safety, Privacy, and Working Responsibly with AI

For professionals:

• Intellectual property: Be deliberate about what proprietary information you put into agent context. Understand your organization’s AI usage policies.
• Regulated industries: If you work in healthcare, legal, finance, or education, know what you can and cannot put into an AI context (HIPAA, attorney-client privilege, FERPA, etc.).
• Client confidentiality: Even if your company allows AI use, your clients may have their own restrictions. Review contracts and NDAs.
• Data residency: Understand where your data goes. Cloud vs local processing, whether data is stored or used for training. Push for enterprise-tier accounts when using AI for work.

For home:

• Don’t put in agent context: your children’s full names, school addresses, medical details, financial information.
• Start conservative – you can always add more context later, but you can’t un-share what’s already been shared.

What Now? Your First Steps

You’ve read the guide. Here’s exactly what to do next, in order. Don’t skip ahead – each step builds on the previous one.

Kickoff Examples

Two complete, ready-to-use agent setups – all six .md files filled in. Download them, read through the files, and use them as a blueprint for your own.

Same structure, completely different domain. Read every line, copy the structure, adapt it for your own use case.

FAQ – Common Questions