Skip to main content
The WebSearchTool provides web search functionality through OpenAI’s agents library. It’s a built-in tool that doesn’t require additional API keys.

​
Class Signature

from agents import WebSearchTool

class WebSearchTool:
    def __init__()

​
Overview

The WebSearchTool is imported from the agents package (part of openai-agents SDK) and provides web search capabilities to Agentor agents. It integrates seamlessly with OpenAI’s agent framework.

​
Usage Example

from agentor import Agentor
from agentor.tools import WebSearchTool

# Create an agent with web search capability
agent = Agentor(
    name="Web Search Agent",
    model="gpt-4",
    tools=[WebSearchTool()],
    instructions="Use web search when up-to-date information is requested.",
)

# Search for current information
result = agent.run("What are the latest developments in AI?")
print(result.final_output)

​
Advanced Example

from agentor import Agentor
from agentor.tools import WebSearchTool

agent = Agentor(
    name="Research Assistant",
    model="gpt-4",
    tools=[WebSearchTool()],
    instructions="""
    You are a research assistant that helps find accurate, 
    up-to-date information. Always cite sources and verify 
    information across multiple search results.
    """,
)

result = agent.run(
    "Find 3 recent updates about Model Context Protocol and summarize each."
)
print(result.final_output)

​
Features

  • No API Key Required: Uses OpenAI’s built-in search capability
  • Real-time Information: Access to current web content
  • Automatic Integration: Works seamlessly with OpenAI’s agent framework
  • Source Attribution: Results typically include source information

​
When to Use

Use WebSearchTool when you need:
  • Current news and events
  • Recent updates on specific topics
  • Real-time data (stock prices, weather, etc.)
  • Information not in the model’s training data
  • Verification of facts and claims

​
Alternative: ExaSearchTool

For more control over web search, consider using ExaSearchTool which provides:
  • Custom search parameters
  • Number of results control
  • Direct API access to Exa search engine
Exa tool docs are not yet published in this docs set.

​
Requirements

# WebSearchTool is available through openai-agents (included in Agentor)
pip install agentor

# Set your OpenAI API key
export OPENAI_API_KEY="your_openai_api_key"

​
Environment Setup

# Required environment variable
export OPENAI_API_KEY="sk-..."

​
Integration with Multiple Tools

from agentor import Agentor
from agentor.tools import WebSearchTool, CalculatorTool, GetWeatherTool

agent = Agentor(
    name="Multi-Tool Assistant",
    model="gpt-4",
    tools=[
        WebSearchTool(),
        CalculatorTool(),
        GetWeatherTool(),
    ],
    instructions="""
    Use the most appropriate tool for each task:
    - Web search for current information
    - Calculator for numerical computations
    - Weather tool for weather data
    """,
)

result = agent.run(
    "Search for the current price of gold, then calculate 10 ounces worth."
)
print(result.final_output)

​
Best Practices

  1. Clear Instructions: Provide clear guidelines on when to use web search
  2. Source Verification: Instruct agents to verify information across multiple sources
  3. Recency Awareness: Use for time-sensitive queries
  4. Fallback Strategy: Combine with other tools for comprehensive capabilities
  5. Rate Limiting: Be mindful of search frequency in production environments

​
Limitations

  • Requires active internet connection
  • Search results depend on OpenAI’s search backend
  • May have rate limits based on your OpenAI plan
  • Search scope and freshness depend on the underlying service
Source: /home/daytona/workspace/source/src/agentor/tools/__init__.py:1
Last modified on March 4, 2026