How to Intermediate · 4 min read

How to define tools for LlamaIndex agent

Q: How to define tools for LlamaIndex agent

To define tools for a LlamaIndex agent, create Python functions or classes that encapsulate specific tasks, then register them as tools with the agent's toolkit. Use the Tool class from llama_index.tools to wrap these functions, enabling the agent to invoke them during execution.

Quick answer

To define tools for a LlamaIndex agent, create Python functions or classes that encapsulate specific tasks, then register them as tools with the agent's toolkit. Use the Tool class from llama_index.tools to wrap these functions, enabling the agent to invoke them during execution.

PREREQUISITES

Python 3.8+
pip install llama-index>=0.6.0
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the llama-index package and set your OpenAI API key as an environment variable.

Install with pip install llama-index openai
Set environment variable in your shell: export OPENAI_API_KEY='your_api_key'

bash

pip install llama-index openai

Step by step

Define a tool as a Python function, wrap it with Tool from llama_index.tools, then create an agent with these tools. The agent can call the tools by name during its reasoning process.

python

import os
from llama_index import GPTSimpleVectorIndex, SimpleDirectoryReader
from llama_index.tools import Tool
from llama_index.agent import Agent

# Define a simple tool function
def greet(name: str) -> str:
    return f"Hello, {name}! Welcome to LlamaIndex agent."

# Wrap the function as a Tool
hello_tool = Tool(
    func=greet,
    name="greet",
    description="Greets a user by name."
)

# Create an agent with the tool
agent = Agent(tools=[hello_tool])

# Use the agent to run a query that triggers the tool
response = agent.chat("Use the greet tool to say hello to Alice.")
print(response)

output

Hello, Alice! Welcome to LlamaIndex agent.

Common variations

You can define tools as classes with a __call__ method for more complex logic. Also, integrate multiple tools by passing a list to the Agent. Use different models by configuring the agent's LLM parameter.

python

from llama_index.tools import Tool
from llama_index.agent import Agent

# Tool as a class
class CalculatorTool:
    def __call__(self, expression: str) -> str:
        try:
            result = eval(expression, {}, {})  # simple eval for demo
            return str(result)
        except Exception:
            return "Invalid expression"

calc_tool = Tool(
    func=CalculatorTool(),
    name="calculator",
    description="Evaluates simple math expressions."
)

agent = Agent(tools=[hello_tool, calc_tool])

response = agent.chat("Calculate 5 * 7 using the calculator tool.")
print(response)

output

Troubleshooting

If the agent does not recognize your tool, ensure the name and description are clear and unique.
For errors during tool execution, add try-except blocks inside your tool functions.
Verify your environment variable OPENAI_API_KEY is set correctly to avoid authentication errors.

✅

Key Takeaways

Use the Tool class from llama_index.tools to define callable tools for the agent.
Register all tools in a list and pass them to the Agent constructor for integration.
Tools can be simple functions or callable classes depending on complexity.
Clear naming and descriptions help the agent select the correct tool during execution.
Always handle exceptions inside tools to prevent agent crashes.

Verified 2026-04 · gpt-4o, llama-3.1-70b

Verify ↗