Module 1: Your First Agent

Learning Objectives

• Understand the Strands Agent abstraction
• Create an agent with a system prompt
• Experience the agentic loop (reason → respond)
• Interact with your first AI agent

Concepts

In this module, we create the simplest possible agent: a model + a system prompt. No tools, no memory, just an LLM that knows how to be a customer support agent.

flowchart LR
    A[💬 User Message] --> B[🧠 System Prompt + Model] --> C[✅ Response]

The system prompt acts as the agent’s “job description”. It defines personality, responsibilities, and constraints.

Hands-On

1. Open the module file

   # From the workshop/ directory
   code module_01_first_agent/agent.py

2. Review the system prompt

   The system prompt defines SupportBot’s behavior:

   SYSTEM_PROMPT = """You are SupportBot, a friendly and helpful
   customer support agent for TechStore..."""

   Key elements of a good system prompt:

   • Role: Who is the agent?
   • Responsibilities: What should it do?
   • Constraints: What should it NOT do?
   • Policies: Company-specific rules

3. Review the agent creation

   from strands import Agent
   from shared.model import model
   
   # Just a model + system prompt - the simplest agent
   agent = Agent(system_prompt=SYSTEM_PROMPT, model=model)

   The model is imported from shared/model.py — the single place where you configure which LLM to use across all modules. See Environment Setup for configuration options.

4. Run the agent

   python module_01_first_agent/agent.py

5. Test with these prompts

   You: Hi there!
   You: What's your return policy?
   You: How long does shipping take?
   You: Can you look up my order?

What’s Happening

When you send a message, the agent:

1. Combines your message with the system prompt
2. Sends it to the LLM configured in shared/model.py
3. The LLM reasons about the best response
4. Streams the response back token by token

Notice that the agent can answer policy questions from its system prompt, but cannot look up specific orders, it has no tools yet. That’s what we’ll add in the next module.

Tip

Try asking the agent to look up an order. It should tell you it can’t. This is the correct behavior since it doesn’t have access to any data tools yet.

Key Takeaways

• An agent is fundamentally: Model + System Prompt + Tools
• The system prompt is critical for agent behavior
• Without tools, the agent can only use knowledge baked into the prompt
• Strands handles the agentic loop, you just define what the agent should be

Next: Adding Tools

In the next module, we’ll give the agent tools to look up orders, search products, and query the FAQ, transforming it from a “smart parrot” to a capable support agent.