Build Your First AI Agent with Gemini – Python Tutorial + GitHub Template

Learn to build your first autonomous AI agent using Gemini and Python. Step-by-step tutorial with GitHub repo, code examples, and diagrams for beginners.

Introduction

Are you curious about building AI agents that can think, act, and manage tasks autonomously? With Gemini and Python, you can create your very own intelligent agent in just a few steps.

In this tutorial, we’ll walk you through:

• Setting up a Gemini AI agent

• Running the agent locally with Python

• Understanding the code and logic behind it

• Linking the ready-to-use GitHub template for easy replication

By the end, you’ll have a functional AI agent capable of managing a simple to-do list and ready for customization.

Prerequisites

Before you start, make sure you have:

• A Gemini subscription

• Python 3.10+ installed

• Basic knowledge of Python programming

• Git installed to clone the repo

Step 1: Clone the GitHub Repository

We’ve prepared a complete template for your AI agent. Clone it locally:

git clone https://github.com/ashish-jabble/ai.git
cd ai/agents/gemini

This repository includes:

• agent_gemini.py → Main agent script
• requirements.txt → Python dependencies
• diagram.png → Visual overview of the agent

---

Step 2: Install Dependencies

Install the required Python packages:

pip install -r requirements.txt

This ensures your agent can communicate with Gemini properly.

---

Step 3: Set Your Gemini API Key

To allow your agent to use Gemini, export your API key as an environment variable.

On macOS / Linux:

export GEMINI_API_KEY="your_key_here"

On Windows (PowerShell):

setx GEMINI_API_KEY "your_key_here"

Important: Never commit your API key to GitHub. Always use environment variables.

---

Step 4: Run Your AI Agent

Start the agent with:

python gemini.py

You will see a prompt like this:

Enter a new task for your AI agent:

Example input:

Find the latest major tech news headlines from the past hour and summarize the top 3 with bullet points.

The agent will display top 3 news from techcrunch.

---

Step 5: Explore the Code

Here’s how the logic works inside gemini.py:

1. Read  bash command and returns the standard output

def execute_bash():
    ...

This function can be used for things like creating directories, moving files, searching, or running curl requests.

3. Gemini Interaction

chat = client.chats.create(model='gemini-2.5-flash', config=config)

---

Step 6: Architecture Diagram

Below is a simple architecture overview:

[User Input] → [Gemini API] → [Agent Logic] → [Output]

• User Input → You provide a new task
• Gemini API → AI processes and prioritizes
• Agent Logic → Python handles task management
• Memory → Tasks saved locally
• Output → Updated prioritized list

---

Step 7: Customize Your Agent

Once your agent works, you can extend it:

• Add multi-step reasoning
• Integrate Google Calendar or email APIs
• Create multi-agent workflows
• Store memory in a database instead of a text file

Gemini Pro allows longer context windows, making it ideal for advanced workflows.

---

Step 8: GitHub Repository

Explore or download the full template here:

View on GitHub → Gemini AI Agent Template

This repository serves as the foundation for building more advanced AI agents using Gemini.

---

Congratulations! 🎉

You have successfully built your first AI agent using Gemini and Python. Now you can extend it, improve it, and integrate it into real-world applications.

Tip: Bookmark this repo — it’s the foundation for all future AI agent experiments!

Conclusion

You’ve now built your first autonomous AI agent using Gemini. With the GitHub template, Python code, and step-by-step tutorial, you can start experimenting with advanced AI workflows.

Share your agent with friends, customize it for your own tasks, and watch your AI skills grow!