Course 01: Your First AI AgentModule 6 of 8

MODULE 06 · COURSE 01: YOUR FIRST AI AGENT

Module 6: Your First Conversation

Your agent is live. Now let's actually use it.


Start simple

Send this:

"Hey, what can you do?"

You'll get a response. It'll probably be pretty generic at this point — because you haven't told your agent who it is or what it's supposed to do yet. That's coming in Module 7.

For now, we're just confirming everything works and getting a feel for the interaction.


Things to try in your first session

Test basic intelligence:
"Explain compound interest like I'm 15."
Test that it remembers within a conversation:
"I run an e-commerce store."
(a few messages later)
"What kinds of things might be useful for my business?"

It should remember what you said about the store.

Test across sessions (memory persistence): End the conversation. Come back tomorrow. Say:
"What do you know about me?"

If memory is working, it'll surface things you've told it. If not — we'll set that up properly.


How conversations work

A few things to understand about how your agent handles messages:

It reads everything you send. Every message you send in a conversation is part of its context. Think of it like a text thread — it can scroll back and see what was said. Context has limits. AI models can only hold so much in memory at once (this is called the "context window"). For most models it's large enough that you won't hit it in normal use. When you do, OpenClaw handles it by compacting older messages. Sessions vs memory. Within one conversation, the agent remembers everything. Across sessions (days later), it relies on its saved MEMORY.md file. This is why the memory setup matters.

What "thinking" looks like

Sometimes you'll ask something complex and there'll be a pause before the response. That's normal — the agent might be:

  • Reasoning through a multi-step problem
  • Using a tool (like searching the web)
  • Processing a long context
Some setups show a typing indicator. Others just go quiet for a moment. Don't resend the message — it's working.

Common first-session mistakes

Treating it like a search engine. Don't ask one question and bounce. Have a conversation. Give it context. "I'm working on X, I'm trying to Y, what should I do about Z?" — that's where it shines. Being too vague. "Help me with my business" is hard to help with. "I run a Shopify store selling handmade candles and I'm trying to write a product description for a new lavender scent" — now it can actually help. Expecting perfection immediately. Your first few interactions will feel a little generic. That's because your agent doesn't know you yet. Every conversation teaches it more.

A tip from Jeremy

When he set up his first agent, he spent the first week just talking to it — explaining his businesses, his preferences, how he thinks. He wasn't asking it to do tasks. He was onboarding it, the same way you'd onboard a new employee.

By week two, it knew enough context that the responses got dramatically better. The time investment pays off.


Next up

Module 7: Give Your Agent a Personality (SOUL.md) →