How to Build a Simple AI Agent in Python for Free (No API Key Needed)

Most AI agent tutorials assume you're paying for the OpenAI API. This one doesn't. You'll build a real, working AI agent in plain Python using Ollama, a free tool that runs open-source language models directly on your own computer. No API key, no billing, no framework.

What Is an AI Agent, Really?

A regular chatbot takes one message and gives one reply. An AI agent is different: it can decide to take an action, look at the result, and decide what to do next, repeating that loop until the task is done. That loop, think, act, observe, repeat, is what separates an agent from a simple prompt-and-response script.

What You'll Need

Python 3.10 or newer
About 8GB of free RAM
Ten minutes, and zero dollars

Step 1: Install Ollama

Ollama lets you download and run open-source models like Llama and Qwen on your own machine, completely free. Go to ollama.com, download the installer for your operating system, and install it like any normal app.

Once installed, pull a small, capable model that supports tool calling:

BASH

ollama pull llama3.1

This downloads the model once, then it runs locally forever with no internet connection and no cost.

Step 2: Set Up Your Project

BASH

mkdir python-ai-agent
cd python-ai-agent
python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate
pip install ollama

That's the only package you need. No paid SDK required.

Step 3: Give Your Agent a Tool

An agent becomes useful the moment it can do something beyond talking, like running a calculation. Here's a simple tool function the agent will be able to call on its own.

PYTHON

def calculate(expression: str) -> str:
    try:
        return str(eval(expression, {"__builtins__": {}}))
    except Exception as e:
        return f"Error: {e}"

Step 4: Write the Agent Loop

This is the core of the tutorial. The agent receives a goal, asks the local model what to do, and either calls the tool or finishes with an answer.

PYTHON

import ollama
 
tools = [
    {
        "type": "function",
        "function": {
            "name": "calculate",
            "description": "Evaluate a basic math expression",
            "parameters": {
                "type": "object",
                "properties": {
                    "expression": {"type": "string"}
                },
                "required": ["expression"],
            },
        },
    }
]
 
def calculate(expression: str) -> str:
    try:
        return str(eval(expression, {"__builtins__": {}}))
    except Exception as e:
        return f"Error: {e}"
 
def run_agent(goal: str):
    messages = [{"role": "user", "content": goal}]
 
    while True:
        response = ollama.chat(
            model="llama3.1",
            messages=messages,
            tools=tools,
        )
        message = response["message"]
        messages.append(message)
 
        if not message.get("tool_calls"):
            return message["content"]
 
        for call in message["tool_calls"]:
            args = call["function"]["arguments"]
            result = calculate(args["expression"])
            messages.append({
                "role": "tool",
                "content": result,
            })
 
if __name__ == "__main__":
    answer = run_agent("What is 42 times 17, plus 100?")
    print(answer)

Step 5: Run It

Make sure Ollama is running in the background (it starts automatically after installation), then run:

BASH

python agent.py

The agent reads your goal, decides it needs the calculate tool, runs it, feeds the result back to itself, and returns a final answer, all running locally on your machine at zero cost.

Why Local Instead of a Paid API?

Running the model locally means unlimited experimentation. You can rerun the agent hundreds of times while you learn, without watching a bill climb. The tradeoff is that a local model is a bit slower and less capable than the largest hosted models, but for learning how agents work, that tradeoff is well worth it.

When You'd Move to a Paid API

Once you're comfortable with the pattern, you might switch to a hosted model for production use, faster responses, or a smarter model on bigger tasks. But there's no reason to pay anything just to learn how agents work.

Final Thoughts

You now have a working AI agent built entirely in free, local Python, no API key required. From here, try swapping in your own tool, a file reader, a to-do list manager, a web scraper, and watch the same loop handle it without any changes to the core logic.

How to Build a Simple AI Agent in Python for Free (No API Key Needed)

How to Build a Simple AI Agent in Python for Free (No API Key Needed)

What Is an AI Agent, Really?

What You'll Need

Step 1: Install Ollama

Step 2: Set Up Your Project

Step 3: Give Your Agent a Tool

Step 4: Write the Agent Loop

Step 5: Run It

Why Local Instead of a Paid API?

When You'd Move to a Paid API

Final Thoughts

Python Functions Explained Like a Recipe (Beginner's Guide)

Discussion & Thoughts

Sign in to leave a comment