AgentPro

AgentPro is a lightweight ReAct-style agentic framework built in Python, designed for structured reasoning step-by-step using available tools, while maintaining a complete history of Thought → Action → Observation → PAUSE → Final Answer steps.

<p align="center"> <img src="https://img.shields.io/badge/Python-3.8%2B-blue" alt="Python 3.8+"> <img src="https://img.shields.io/badge/License-Apache%202.0-blue" alt="License: Apache 2.0"> </p>

📚 Features

• 🔥 ReAct (Reasoning + Acting) agent pattern
• 🛠️ Modular tool system (easy to add your own tools)
• 📜 Clean Thought/Action/Observation/PAUSE parsing
• 📦 Local package structure for easy extension
• 🧠 Powered by any LLM! (Anthropic, Open AI or any other Open source LLMs)

Quick Start

Installation

Install agentpro repository using pip:

pip install git+https://github.com/traversaal-ai/AgentPro.git -q

<!-- ### Configuration Create a `.env` file in the root directory with your API keys: ``` OPENAI_API_KEY=your_openai_api_key TRAVERSAAL_ARES_API_KEY=your_traversaal_ares_api_key ``` Ares internet tool: Searches the internet for real-time information using the Traversaal Ares API. To get `TRAVERSAAL_ARES_API_KEY`. Follow these steps: 1. Go to the [Traversaal API platform](https://api.traversaal.ai/) 2. Log in or create an account 3. Click **"Create new secret key"** 4. Copy the generated key and paste in `.env` file : ### Running the Agent From the command line: ```bash python main.py ``` This starts an interactive session with the agent where you can enter queries. -->

Basic Usage

import os
from agentpro import ReactAgent
from agentpro.tools import AresInternetTool
from agentpro import create_model

# Create a model with OpenAI
model = create_model(provider="openai", model_name="gpt-4o", api_key=os.getenv("OPENAI_API_KEY", None))

# Initialize tools
tools = [AresInternetTool(os.getenv("ARES_API_KEY", None))]

# Initialize agent
agent = ReactAgent(model=model, tools=tools)

# Run a query
query = "What is the height of the Eiffel Tower?"
response = agent.run(query)

print(f"\nFinal Answer: {response.final_answer}")

For Ares api key, follow these steps:

1. Go to the Traversaal API platform
2. Log in or create an account
3. Generate your Ares API key from the dashboard.

<!-- You can also use the [Quick Start](https://github.com/traversaal-ai/AgentPro/blob/main/cookbook/quick_start.ipynb) Jupyter Notebook to run AgentPro directly in Colab. -->

🌍 Traversaal x Optimized AI Hackathon 2025

We’re teaming up with the Optimized AI Conference 2025 to host a global hackathon on AI Agents — open to all developers, builders, researchers, and dreamers working on intelligent systems.

The Challenge

Build a real, functional AI Agent that solves a real-world problem.

This isn’t about flashy demos. We want to see domain-specific, usable, vertical agents — like:

• 🧑‍💼 Customer Support Agents
• 🔬 Research Assistants
• 📊 Data Analyst Agents
• 💡 Or something totally original

You can use any framework, but we recommend trying AgentPro — our open-source toolkit designed for rapid prototyping and robust architecture.

Key Dates

• Hackathon Starts: April 9, 2025
• Submission Deadline: April 15, 2025
• Winners Announced: April 15, 2025 (Live @ Optimized AI Conference)

Prizes + Recognition

| Prize Tier | Reward | |--------------------|------------| | 🥇 Grand Prize | $1,000 | | 🥈 Runner-Up | $500 | | 🥉 Honorable Mention x2 | $250 |

Plus:

• 1:1 Mentorship opportunities
• Invitation to Traversaal’s AI Fellowship Program

Want to be a Judge?

We’re looking for global experts in AI, product, UX, and enterprise applications to help evaluate the submissions. 👉 Apply to be a Judge

For more details, follow this link

📩 Questions? Reach us at hackathon-oai@traversaal.ai

🛠️ Creating Custom Tools

You can create your own tools by extending the Tool base class provided in agentpro.

Here’s a basic example:

from agentpro.tools import Tool
from typing import Any

class MyCustomTool(Tool):
    name: str = "My Custom Tool"  # Human-readable name for the tool (used in documentation and debugging)
    description: str = "Description"  # Brief summary explaining the tool's functionality for agent
    action_type: str = "my_custom_action"  # Unique identifier for the tool; lowercase with underscores for agent; avoid spaces, digits, special characters
    input_format: str = "Description of expected input format, e.g., a string query."  # Instruction on what kind of input the tool expects with example

    def run(self, input_text: Any) -> str:
        # your tool logic
        return "Result of your custom tool."

After creating your custom tool, you can initialize it and pass it to AgentPro like this:

import os
from agentpro import ReactAgent
from agentpro import create_model

# Create a model with OpenAI
model = create_model(provider="openai", model_name="gpt-4o", api_key=os.getenv("OPENAI_API_KEY", None))

# Instantiate your custom tools
tools = [MyCustomTool()]

# Create AgentPro React agent
myagent = ReactAgent(model=model,tools=tools)

# Run a query
query = "Use the custom tool to perform a task."
response = myagent.run(query)

print(response.final_answer)

Project Structure

AgentPro/
├── agentpro/
│   ├── __init__.py
│   ├── react_agent.py                  # Core AgentPro class implementing react-style agent framework
│   ├── agent.py                        # Action, Observation, ThoughtStep, AgentResponse classes
│   ├── model.py                        # Model classes 
│   ├── tools/                          # folder for all tool classes
│       ├── __init__.py
│       ├── base_tool.py
│       ├── duckduckgo_tool.py
│       ├── calculator_tool.py
│       ├── userinput_tool.py
│       ├── ares_tool.py
│       ├── traversaalpro_rag_tool.py
│       ├── slide_generation_tool.py
│       └── yfinance_tool.py
├── cookbook/
│   ├── Traversaal x Optimized AI Hackathon 2025
│   ├── quick_start.ipynb
│   └── custool_tool.ipynb      
├── main.py                             # Entrypoint to run the agent
├── requirements.txt                    # Dependencies
├── README.md                           # Project overview, usage instructions, and documentation
├── setup.py       
├── pyproject.toml     
└── LICENSE.txt                         # Open-source license information (Apache License 2.0)

Requirements

• Python 3.8+
• OpenAI API key
• Traversaal Ares API key for internet search (Optional)

License

This project is licensed under the Apache 2.0 License - see the LICENSE file for more details.