Building Your First AI Agent: A Hands-on Guide

Photo by Hung Nguyen Phi on Unsplash

Hey there! 👋 If you’re joining us from the first article, welcome back! If you’re new here, no worries — just check out the first article to get up to speed.

Today, we’re rolling up our sleeves and building something cool: a smart agent that can actually learn from its environment. No more “if-this-then-that” stuff — we’re going full autonomous!

What We’re Building

We’re creating a trading agent that can:
- Read market data
- Make decisions based on patterns
- Learn from its successes and failures
- Adapt its strategy over time

Don’t worry if you’ve never done anything with trading — the principles we’ll learn apply to any kind of agent system.

The Sense-Plan-Act Cycle

Remember our digital butler analogy from last time? Let’s upgrade it. Every smart agent follows what we call the Sense-Plan-Act cycle. Think of it like a human making decisions:

1. Sense: Get information (like checking your phone’s weather app)
2. Plan: Process that information (like deciding whether to take an umbrella)
3. Act: Do something based on your plan (like actually grabbing that umbrella)

Here’s how we’ll implement this cycle:

from dataclasses import dataclass
from typing import List, Dict
import numpy as np
@dataclass
class MarketState:
 price: float
 volume: float
 trend: str # 'up', 'down', or 'sideways'
class TradingAgent:
 def __init__(self):
 self.memory = [] # Store past observations
 self.performance = 0 # Track how well we're doing
 self.position = None # Current trading position
 
 def sense(self, market_data: MarketState) -> Dict:
 """
 Process raw market data into useful information
 """
 observation = {
 'price': market_data.price,
 'volume': market_data.volume,
 'trend': market_data.trend,
 'price_change': self._calculate_price_change(),
 'volatility': self._calculate_volatility()
 }
 self.memory.append(observation)
 return observation
 
 def plan(self, observation: Dict) -> str:
 """
 Decide what action to take based on current observation
 """
 if len(self.memory) < 5: #…