Build a Simple AI Agent with Python and CrewAI

TL;DR: Build a Python AI agent using CrewAI, Serper, and Google 2.0 Flash APIs. The agent researches a topic and writes a blog post.
📦 GitHub Repo

Introduction

In this guide, you'll learn how to build a basic AI agent using Python and CrewAI. The agent will search the internet and generate a blog summary using the free Google 2.0 Flash and Serper APIs. This beginner-friendly tutorial walks you through each step.

Getting Started

Ensure Python is installed on your system. You can download it from the official website.

Create a new project directory:

mkdir simple-ai-agent

cd simple-ai-agent

Create a requirements.txt file and add the necessary packages:

Install the dependencies:

pip install -r requirements.txt

Create the main Python file:

touch agent.py

Set Up

Open agent.py and add the initial setup code:

This code uses:

• CrewAI : A library for creating AI agents.
• Serper : An API for searching the internet.
• Google 2.0 Flash : A free LLM API for generating text.

Create a .env file and add your API keys:

Load the environment variables in agent.py:

Define Tools and LLM

Define the tools and LLM for the agent:

from crewai_tools import SerperDevTool
from langchain_google_genai import GoogleGenerativeAI

search_tool = SerperDevTool()

llm = GoogleGenerativeAI(model="gemini-2.0-flash",google_api_key=GOOGLE_AI_API_KEY)

User Input

Prompt the user to enter a topic:

print("Welcome to AI Researcher and Writer!")

topic = input("Enter the topic: ") # Prompt user to enter a topic

Creating the Agents

Create two agents: a researcher and a writer.

researcher = Agent(
    role="Researcher",
    goal=f"Uncover interesting findings about {topic}",
    verbose=True,
    memory=True,
    backstory=(
    """
    As a researcher, you are committed to uncovering the
    latest and most interesting findings in your field. You
    have a knack for finding hidden gems of information and
    presenting them in an engaging way. Your goal is to
    illuminate the topic at hand, providing insights that are
    both informative and thought-provoking.
    """
    ),
    tools=[search_tool],
    llm=llm,
    allow_delegation=True
)

writer = Agent(
    role="Writer",
    goal=f"Write intuitive article about {topic}",
    verbose=True,
    memory=True,
    backstory=(
    """
    As a writer, you are dedicated to crafting engaging
    and informative articles. You have a talent for
    transforming complex ideas into accessible language,
    making them relatable to a wide audience.
    """
    ),
    tools=[search_tool],
    llm=llm,
    allow_delegation=False
)

Creating the Tasks

Define tasks for each agent:

research_task = Task(
    description=(
        f"Drive key insights about {topic}."
        "What are the latest trends, technologies, and innovations?"
        "Have a balanced view, considering both the positive and negative aspects."
        "Your report should be well-structured and easy to follow."
    ),
    expected_output="A comprehensive 3 paragraphs long report on the topic",
    tools=[search_tool],
    agent=researcher
)

write_task = Task(
    description=(
        f"Compose an detailed and easy to understand article on {topic}"
        "The article should be engaging and informative, suitable for a general audience."
        "It should be well-structured, with a clear introduction, body, and conclusion."
        "Use markdown formatting for headings, lists, and code snippets where appropriate."
        "The article should be at least 4 paragraphs long and cover the key points from the research report."
    ),
    expected_output=f"A 4 paragraph article on {topic} fomatted as markdown",
    tools=[search_tool],
    agent=writer,
    async_execution=False,
    output_file="blog-post.md"
)

Running the Tasks

Run the agent crew sequentially:

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential
)

result = crew.kickoff()
print(result)

Your agent.py should now resemble:

import os
from dotenv import load_dotenv
from crewai import Agent, Task, Crew, Process
from crewai_tools import SerperDevTool
from langchain_google_genai import GoogleGenerativeAI

load_dotenv() # Load environment variables from .env file

SERPER_API_KEY = os.getenv("SERPER_API_KEY")
GOOGLE_AI_API_KEY = os.getenv("GOOGLE_AI_API")

search_tool = SerperDevTool()

llm = GoogleGenerativeAI(model="gemini-2.0-flash",google_api_key=GOOGLE_AI_API_KEY)

print("Welcome to AI Researcher and Writer!")
topic = input("Enter the topic: ") # Prompt user to enter a topic

researcher = Agent(
    role="Researcher",
    goal=f"Uncover interesting findings about {topic}",
    verbose=True,
    memory=True,
    backstory=(
    """
    As a researcher, you are committed to uncovering the
    latest and most interesting findings in your field. You
    have a knack for finding hidden gems of information and
    presenting them in an engaging way. Your goal is to
    illuminate the topic at hand, providing insights that are
    both informative and thought-provoking.
    """
    ),
    tools=[search_tool],
    llm=llm,
    allow_delegation=True
)

writer = Agent(
    role="Writer",
    goal=f"Write intuitive article about {topic}",
    verbose=True,
    memory=True,
    backstory=(
    """
    As a writer, you are dedicated to crafting engaging
    and informative articles. You have a talent for
    transforming complex ideas into accessible language,
    making them relatable to a wide audience.
    """
    ),
    tools=[search_tool],
    llm=llm,
    allow_delegation=False
)

research_task = Task(
    description=(
        f"Drive key insights about {topic}."
        "What are the latest trends, technologies, and innovations?"
        "Have a balanced view, considering both the positive and negative aspects."
        "Your report should be well-structured and easy to follow."
    ),
    expected_output="A comprehensive 3 paragraphs long report on the topic",
    tools=[search_tool],
    agent=researcher
)

write_task = Task(
    description=(
        f"Compose an detailed and easy to understand article on {topic}"
        "The article should be engaging and informative, suitable for a general audience."
        "It should be well-structured, with a clear introduction, body, and conclusion."
        "Use markdown formatting for headings, lists, and code snippets where appropriate."
        "The article should be at least 4 paragraphs long and cover the key points from the research report."
    ),
    expected_output=f"A 4 paragraph article on {topic} fomatted as markdown",
    tools=[search_tool],
    agent=writer,
    async_execution=False,
    output_file="blog-post.md"
)

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential
)

result = crew.kickoff()
print(result)

Execute the code:

python agent.py

The generated blog post will be saved as blog-post.md in the project directory. Open it in any text editor.

You can find the demo code on Github Repo