MODULE 3 OF 10

Teaching Your Agent About You

Build memory so your agent learns your preferences, patterns, and important context over time

🧠 The Three Types of Memory

Your agent has three types of memory that work together to create a personalized experience. Understanding how each works helps you teach your agent more effectively.

⚡

Short-Term

Active conversation context. What you're discussing right now.

Current topic
Recent messages
Thread context

📋

Mid-Term

Session memory. Things you've taught during this conversation.

Preferences mentioned
Decisions made
Context shared

💾

Long-Term

Persistent memory files. USER.md, MEMORY.md, and custom files.

Your profile
Past learnings
Reference docs

💡

Pro Tip

Explicitly telling your agent "remember this" or "add this to my profile" moves information from short/mid-term into long-term memory files automatically!

🎯 Example 1: Teaching Preferences

You

I prefer my emails to be concise and professional. No emojis, straight to the point.

Your Agent

Got it! I'll remember that you prefer concise, professional communication without emojis. I'll add this to your USER.md profile so all future interactions follow this style.

You (later)

Draft an email to my team about the project delay.

Your Agent

[Drafts concise, professional email without emojis, referencing your preference]

🎯 Example 2: Building Context Over Time

You (Week 1)

I'm allergic to shellfish and prefer vegetarian options when dining out.

Your Agent

Noted! Shellfish allergy + vegetarian preference. I'll remember this for restaurant recommendations.

You (Week 3)

Find me a good restaurant for dinner tonight near my hotel.

Your Agent

Here are 5 vegetarian-friendly restaurants near your hotel, all confirmed shellfish-free. I've prioritized places with diverse vegetarian menus based on your preferences.

🎯 Example 3: Reference Files

You

Save this company style guide for future reference: [pastes brand guidelines]

Your Agent

I'll create a reference file called COMPANY_STYLE_GUIDE.md with these guidelines. I can reference this whenever you need content that matches your brand voice.

You (months later)

Write a blog post announcing our new feature.

Your Agent

[Writes blog post following your company's style guide, tone, and formatting preferences from the saved reference file]

👤 Building Your USER.md

USER.md is your agent's cheat sheet about you. It's the foundation of long-term memory. The more complete it is, the better your agent can anticipate your needs.

# USER.md - Your Personal Profile

Name: Alex Chen
Timezone: America/New_York (EST/EDT)
Location: Brooklyn, NY

## Communication Preferences
Email style: Concise, professional, no emojis
Meeting preference: Async when possible, 30min max
Urgency signals: "URGENT" in subject = drop everything

## Work Context
Role: Product Manager at TechCorp
Team size: 8 engineers, 2 designers
Current focus: Q3 roadmap, mobile app redesign

## Personal Context
Diet: Vegetarian, allergic to shellfish
Interests: Hiking, photography, sci-fi books
Family: Partner Sarah, dog Max
            

🧩 Quick Quiz

Which type of memory persists between conversations?

⚡ Short-term memory

📋 Mid-term memory

💾 Long-term memory (files like USER.md)

✏️ Practice Exercises

Create Your USER.md Foundation

Start building your USER.md with the basics. Copy this prompt and fill in your details:

Create a USER.md file for me with the following sections: - Basic info (name, timezone, location) - Communication preferences (email style, meeting preferences) - Work context (role, team, current projects) - Personal context (diet, interests, important people) Here's my info: [add your details here]

Teach a Preference

Practice moving information from conversation to long-term memory:

Remember this preference: I prefer [your preference here]. Add this to my USER.md under Communication Preferences.

Create a Reference File

Save something important for future reference:

Create a reference file called [TOPIC]_REFERENCE.md with this information: [paste important info, guidelines, or context here] I'll want to reference this in future conversations.

🎉

Module 3 Complete!

You now understand how your agent remembers things. Next up: exploring everything your agent can do.