What's New in CrewAI - Expanded Toolset and Integrations

Lennex Zinyando

• 5 min read

The fast-paced world of AI requires tools that move just as fast. CrewAI has a reputation for shipping new features and bug fixes at the speed of light. In the last few months, the CrewAI team has shipped a LOT of releases and it might be hard for the ordinary developer to keep track of all the new features available to them. This blog post series highlights some of these new features available to you introduced between v0.30.0 and v0.36.0. Since there is a lot to talk about I have broken down the blog posts into 3 posts

  • Expanded toolset and integrations ← this post
  • Enhanced execution and control
  • Agent Management and Customization

New Native Tools

Code Execution by Agents

CrewAI introduced a CodeInterpreterTool that enables AI agents to execute Python3 code within a safe, sandboxed environment. It’s important to note that the code must be generated by the agent itself for it to be executed. Agents can generate code, execute it, obtain results, and use these results for decision-making. This system is based on Docker so initial execution may be slower due to Docker image building.

pip install 'crewai[tools]'

from crewai import Agent
from crewai_tools import CodeInterpreterTool

Agent(
    ...
    tools=[CodeInterpreterTool()],
)

This tool significantly enhances an AI agent's capabilities by allowing it to dynamically create and run code as part of its problem-solving process.

Browserbase

Browserbase is a serverless developer platform designed to manage and operate headless browsers at scale. It's handy for web automation and AI agent tasks. Key features include:

  1. Compatibility with popular browser automation tools like Playwright, Puppeteer, and Selenium.
  2. Integration with AI frameworks, including CrewAI and Langchain.
  3. A Session Inspector for full observability of browser sessions.
  4. Stealth mode with automatic captcha solving and residential proxies.
  5. Support for advanced features such as custom extensions, file downloads, and long-running sessions.
  6. APIs for integrating live views and retrieving logs and recordings.

Browserbase aims to provide a secure, scalable, and reliable infrastructure for developers working with headless browsers, offering transparent pricing and robust support for various web automation tasks.

To use use the browserbase tool you need to install the the Browserbase SDK along with crewai[tools] package

pip install browserbase 'crewai[tools]'

from crewai_tools import BrowserbaseLoadTool

tool = BrowserbaseLoadTool()

Exa Search is an advanced, neural search engine designed specifically for use with Large Language Models (LLMs) and AI applications. Here are Exa Search's key features in 3:

  1. Neural-based search engine: Uses transformer models to understand query context and meaning, providing more relevant results than traditional keyword-based searches.
  2. LLM-optimized: Designed to integrate with and enhance Large Language Models by providing up-to-date, high-quality web content, reducing hallucinations and improving accuracy.
  3. Advanced capabilities: Offers natural language query processing, instant content retrieval, and similarity search based on content and style, with frequent index updates every two minutes.

To use the EXASearchTool you need to install crewai[tools]

pip install 'crewai[tools]'

from crewai_tools import EXASearchTool

# Initialize the tool for internet searching capabilities
tool = EXASearchTool()

Firecrawl Tools

Firecrawl is a web scraping and data extraction tool designed specifically for AI and machine learning applications. It offers:

  1. Website-to-markdown conversion: Transforms entire websites into clean, LLM-ready markdown or structured data formats.
  2. Simplified data acquisition: Provides a single API for scraping, crawling, and extracting web data, streamlining the process for AI companies and developers.
  3. AI-focused functionality: Tailored for LLM engineers, data scientists, and AI researchers, facilitating easier data preparation for training models, conducting market research, and aggregating content.

CrewAI offers 3 tools that you can use with FireCrawl ie FirecrawlCrawlWebsiteTool, FirecrawlScrapeWebsiteTool and FirecrawlSearchTool . You also need to get an API key from FireCrawl to be able to use the tools.

pip install firecrawl-py 'crewai[tools]'

from crewai_tools import FirecrawlSearchTool, FirecrawlScrapeWebsiteTool, FirecrawlCrawlWebsiteTool

tool = FirecrawlSearchTool(query='what is firecrawl?')

tool = FirecrawlScrapeWebsiteTool(url='firecrawl.dev')

tool = FirecrawlCrawlWebsiteTool(url='firecrawl.dev')

Third party tool integrations

CrewAI offers integration with tools from other frameworks namely LangChain, LlamaHub and Composio. Allowing for such integrations has some benefits namely

  1. Expanded Functionality:
    • These integrations allow CrewAI users to access a wider range of tools and capabilities without having to switch between different frameworks.
    • Each framework brings its own unique set of features, enriching CrewAI's offerings.
  2. Ecosystem Compatibility:
    • By integrating with popular frameworks, CrewAI ensures compatibility with existing AI ecosystems, making it easier for developers to adopt and incorporate into their workflows.
    • This compatibility reduces friction for teams already using tools from these other frameworks.
  3. Specialization and Expertise:
    • Different frameworks often specialize in particular areas. By integrating them, CrewAI can leverage their specific strengths without having to reinvent the wheel.
    • For example, LangChain is known for its language model tools, while LlamaHub offers a variety of AI-powered integrations.
  4. Time and Resource Efficiency:
    • Developers can save significant time and resources by using pre-built, well-tested tools from these frameworks instead of developing similar functionalities from scratch.
    • This efficiency allows teams to focus on their core objectives rather than building supporting infrastructure.
  5. Community and Support:
    • Integration with popular frameworks means CrewAI users can benefit from larger community support, documentation, and resources associated with these tools.
    • This can lead to faster problem-solving and more robust implementations.
  6. Future-Proofing:
    • As the AI landscape evolves, these integrations allow CrewAI to stay current by easily incorporating new advancements from partner frameworks.
    • This ensures that CrewAI remains a relevant and powerful tool in the rapidly changing field of AI development.

By offering these third-party integrations, CrewAI positions itself as a flexible, powerful, and user-friendly framework that can adapt to various development needs and preferences in the AI community.

AgentOps Native Support

AgentOps is a monitoring and observability platform designed specifically for AI agents and Large Language Model (LLM) applications. It is a tool that provides real-time monitoring, logging, and analytics for AI agents and LLM-powered applications. It's designed to help developers and teams gain insights into their AI systems' performance and behavior.

What it is used for:

  1. Monitoring AI agent interactions and performance
  2. Tracking and analyzing LLM usage and costs
  3. Debugging and optimizing AI workflows
  4. Visualizing agent decision-making processes

Benefits of having a native integration

  1. Seamless Monitoring: Provides real-time insights into AI agent performance directly within the CrewAI ecosystem, eliminating the need to switch between tools.
  2. Enhanced Debugging: Allows developers to quickly identify and resolve issues specific to CrewAI's architecture and agent interactions, leading to faster development cycles.
  3. Resource Optimization: Built-in cost tracking helps teams manage and optimize their LLM usage within CrewAI projects, potentially reducing operational costs.
  4. Tailored Insights: Offers analytics and visualizations specifically designed for CrewAI's architecture, providing more relevant and actionable insights for developers.
  5. Streamlined Workflow: Keeps developers within the CrewAI environment for both development and monitoring, improving productivity and easing the adoption of best practices in AI system observability.

AgentOps aims to bridge the gap between traditional software monitoring tools and the unique needs of AI and LLM applications, providing developers with the tools they need to build more reliable and efficient AI systems.

CrewAI has grown a lot with its updates from versions 0.30 to 0.36. It now has new tools like Code Interpreter, Browserbase, Exa Search, and Firecrawl built right in. It also works better with other AI tools from other ecosystems like LangChain, LlamaHub and Composio, and even keeps an eye on how your AI is doing with AgentOps. All of this means CrewAI can now do many more things, from running code to searching the web in smart ways.

As AI keeps changing, CrewAI is changing too, making sure it stays useful for people building AI agents. These new features help developers work faster and make better AI systems. With all these improvements, CrewAI is helping shape how we'll use AI in the future, giving developers the tools to create smarter and more helpful AI agents.

AI should drive results, not complexity. AgentemAI helps businesses build scalable, efficient, and secure AI solutions. See how we can help.