CrewAI Tools

Warnings

• breaking Changes within the core CrewAI framework (e.g., in output types for tasks/crews, event system refactors) can indirectly affect how tools are integrated and how their outputs are processed. Always review CrewAI's changelog when upgrading `crewai-tools`.
  Fix: Consult the official CrewAI documentation and release notes for migration guides. Adapt agent and task definitions to align with new CrewAI API expectations, especially regarding `TaskOutput` and `CrewOutput` types.
• gotcha Many powerful tools (e.g., `SerperDevTool`, `GithubSearchTool`, various RAG tools) require API keys or specific credentials to be set as environment variables (e.g., `SERPER_API_KEY`, `GITHUB_PAT`) or passed directly. Failure to configure these will lead to runtime errors.
  Fix: Refer to the specific tool's documentation for required environment variables or initialization parameters. Ensure all necessary API keys are securely configured before execution.
• deprecated CVEs (Common Vulnerabilities and Exposures) in underlying dependencies (e.g., `cryptography`, `transformers`, `litellm`) are frequently patched in `crewai-tools` releases. Running outdated versions may expose your application to known security risks.
  Fix: Regularly update `crewai-tools` to the latest stable version. Monitor release notes for security fixes and dependency updates to ensure your environment is protected.
• gotcha When creating custom tools, if your tool involves asynchronous operations (e.g., network requests), you must implement the `_arun` method instead of the synchronous `_run` method. Using `_run` with `async` logic will lead to blocking behavior or errors.
  Fix: For asynchronous custom tools, subclass `BaseTool` and implement `async def _arun(self, *args, **kwargs): ...`. Ensure `crewai` agents and tasks are configured for asynchronous execution if applicable.

Install

• pip install crewai-tools Install stable version

Imports

• FileReadTool

  from crewai_tools import FileReadTool

• SerperDevTool

  from crewai_tools import SerperDevTool

• ScrapeWebsiteTool

  from crewai_tools import ScrapeWebsiteTool

Quickstart

This quickstart demonstrates how to initialize and use two common `crewai-tools`: `FileReadTool` and `SerperDevTool`. It shows how to assign these tools to `crewai` agents, define tasks for the agents, and then run a simple sequential crew to perform research and content creation. Note that `SERPER_API_KEY` must be set for `SerperDevTool` to function, and `crewai` must also be installed.

import os
from crewai import Agent, Task, Crew, Process
from crewai_tools import FileReadTool, SerperDevTool

# Set your Serper API key as an environment variable or pass directly
# os.environ["SERPER_API_KEY"] = "YOUR_SERPER_API_KEY"

# Initialize tools
read_tool = FileReadTool(file_path='./my_document.txt')
search_tool = SerperDevTool()

# Define Agents
researcher = Agent(
    role='Senior Researcher',
    goal='Uncover the latest trends in AI and provide a summary',
    backstory='An expert in AI research, known for insightful analysis.',
    verbose=True,
    allow_delegation=False,
    tools=[search_tool] # Agent has access to the search tool
)

writer = Agent(
    role='Content Writer',
    goal='Craft engaging content based on research findings',
    backstory='A talented writer who transforms complex data into compelling narratives.',
    verbose=True,
    allow_delegation=False,
    tools=[read_tool] # Agent has access to the file read tool
)

# Define Tasks
research_task = Task(
    description='Conduct a comprehensive search on current AI trends and identify key developments.',
    expected_output='A bullet-point summary of 3-5 major AI trends.',
    agent=researcher
)

write_task = Task(
    description='Write a short blog post (approx. 300 words) based on the AI trends summary provided.',
    expected_output='A well-structured and engaging blog post.',
    agent=writer
)

# Assemble a Crew
crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential,
    verbose=2
)

# Ensure 'my_document.txt' exists for the FileReadTool example
with open('./my_document.txt', 'w') as f:
    f.write('Initial context about AI: AI is rapidly advancing, with new models and applications emerging constantly.')

# Kick off the crew's work
print("Crew starting to work...")
result = crew.kickoff()
print("\n\nCrew finished with results:")
print(result)