Crawl4AI - AG2

AG2 now integrates Crawl4AI, an open-source web crawler built for AI agents, language models, and data pipelines. This allows your agents to quickly access and process web data in real time, making automation and data collection more efficient.

Installation

To integrate Crawl4AI with AG2, follow these steps:

Install AG2 with the crawl4ai extra:
```
pip install ag2[openai,crawl4ai]
```
If you have been using autogen or ag2, all you need to do is upgrade it using:
pip install -U autogen[openai,crawl4ai]
or
pip install -U ag2[openai,crawl4ai]
as autogen and ag2 are aliases for the same PyPI package.

Set up Playwright:

# Installs Playwright and browsers for all OS
playwright install
# Additional command, mandatory for Linux only
playwright install-deps

For running the code in Jupyter, use nest_asyncio to allow nested event loops.
```
pip install nest_asyncio
```

Once installed, you’re ready to start using the browsing features in AG2.

Imports

import os

import nest_asyncio
from pydantic import BaseModel

from autogen import AssistantAgent, UserProxyAgent, LLMConfig
from autogen.tools.experimental import Crawl4AITool

nest_asyncio.apply()

Agent Configuration

Configure the agents for the interaction.

config_list defines the LLM configurations, including the model and API key.
UserProxyAgent simulates user inputs without requiring actual human interaction (set to NEVER).
AssistantAgent represents the AI agent, configured with the LLM settings.

Crawl4AI is built on top of LiteLLM and supports the same models as LiteLLM.We had great experience with OpenAI, Anthropic, Gemini and Ollama. However, as of this writing, DeepSeek is encountering some issues.

config_list = [{"}]

llm_config = LLMConfig(api_type="openai", model="gpt-4o-mini", api_key=os.environ["OPENAI_API_KEY"])

user_proxy = UserProxyAgent(name="user_proxy", human_input_mode="NEVER")
with llm_config:
    assistant = AssistantAgent(name="assistant")

Web Browsing with `Crawl4AI`

Crawl4AITool offers three integration modes:

Basic Web Scraping (No LLM):

crawlai_tool = Crawl4AITool()

Web Scraping with LLM Processing:

crawlai_tool = Crawl4AITool(llm_config=llm_config)

LLM Processing with Schema for Structured Data:

class Blog(BaseModel):
    title: str
    url: str


crawlai_tool = Crawl4AITool(llm_config=llm_config, extraction_model=Blog)

We’ll proceed with the most advanced option: LLM Processing with Structured Data Schema. After creating the tool instance, register the agents:

crawlai_tool.register_for_execution(user_proxy)
crawlai_tool.register_for_llm(assistant)

Initiate Chat

message = "Extract all blog posts from https://docs.ag2.ai/docs/blog"
result = user_proxy.initiate_chat(
    recipient=assistant,
    message=message,
    max_turns=2,
)

user_proxy (to assistant):

Extract all blog posts from https://docs.ag2.ai/docs/blog

--------------------------------------------------------------------------------
INFO:httpx:HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
assistant (to user_proxy):

***** Suggested tool call (call_R1DUz9LKPxE3t0wiy94OotAR): crawl4ai *****
Arguments:
{"url":"https://docs.ag2.ai/docs/blog","instruction":"Extract all blog posts including titles, dates, and links."}
*************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION crawl4ai...
Call ID: call_R1DUz9LKPxE3t0wiy94OotAR
Input arguments: {'url': 'https://docs.ag2.ai/docs/blog', 'instruction': 'Extract all blog posts including titles, dates, and links.'}
[INIT].... → Crawl4AI 0.4.247
[FETCH]... ↓ https://docs.ag2.ai/docs/blog... | Status: True | Time: 8.28s
[SCRAPE].. ◆ Processed https://docs.ag2.ai/docs/blog... | Time: 23ms
12:16:26 - LiteLLM:INFO: utils.py:2825 -
LiteLLM completion() model= gpt-4o-mini; provider = openai
INFO:LiteLLM:
LiteLLM completion() model= gpt-4o-mini; provider = openai
INFO:httpx:HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
12:17:04 - LiteLLM:INFO: utils.py:1030 - Wrapper: Completed Call, calling success_handler
INFO:LiteLLM:Wrapper: Completed Call, calling success_handler
[EXTRACT]. ■ Completed for https://docs.ag2.ai/docs/blog... | Time: 38.029349499964155s
[COMPLETE] ● https://docs.ag2.ai/docs/blog... | Status: True | Total: 46.34s
user_proxy (to assistant):

***** Response from calling tool (call_R1DUz9LKPxE3t0wiy94OotAR) *****
[
    {
        "title": "RealtimeAgent with Gemini API",
        "url": "https://docs.ag2.ai/docs/blog/2025-01-29-RealtimeAgent-with-gemini/index",
        "error": false
    },
    {
        "title": "Tools with ChatContext Dependency Injection",
        "url": "https://docs.ag2.ai/docs/blog/2025-01-22-Tools-ChatContext-Dependency-Injection/index",
        "error": false
    },
    {
        "title": "Streaming input and output using WebSockets",
        "url": "https://docs.ag2.ai/docs/blog/2025-01-10-WebSockets/index",
        "error": false
    },
    {
        "title": "Real-Time Voice Interactions over WebRTC",
        "url": "https://docs.ag2.ai/docs/blog/2025-01-09-RealtimeAgent-over-WebRTC/index",
        "error": false
    },
    {
        "title": "Real-Time Voice Interactions with the WebSocket Audio Adapter",
        "url": "https://docs.ag2.ai/docs/blog/2025-01-08-RealtimeAgent-over-websocket/index",
        "error": false
    },
    {
        "title": "Tools Dependency Injection",
        "url": "https://docs.ag2.ai/docs/blog/2025-01-07-Tools-Dependency-Injection/index",
        "error": false
    },
    {
        "title": "Cross-Framework LLM Tool Integration with AG2",
        "url": "https://docs.ag2.ai/docs/blog/2024-12-20-Tools-interoperability/index",
        "error": false
    },
    {
        "title": "ReasoningAgent Update - Beam Search, MCTS, and LATS for LLM Reasoning",
        "url": "https://docs.ag2.ai/docs/blog/2024-12-20-Reasoning-Update/index",
        "error": false
    },
    {
        "title": "Introducing RealtimeAgent Capabilities in AG2",
        "url": "https://docs.ag2.ai/docs/blog/2024-12-20-RealtimeAgent/index",
        "error": false
    },
    {
        "title": "Knowledgeable Agents with FalkorDB Graph RAG",
        "url": "https://docs.ag2.ai/docs/blog/2024-12-06-FalkorDB-Structured/index",
        "error": false
    },
    {
        "title": "ReasoningAgent - Tree of Thoughts with Beam Search in AG2",
        "url": "https://docs.ag2.ai/docs/blog/2024-12-02-ReasoningAgent2/index",
        "error": false
    },
    {
        "title": "Agentic testing for prompt leakage security",
        "url": "https://docs.ag2.ai/docs/blog/2024-11-27-Prompt-Leakage-Probing/index",
        "error": false
    },
    {
        "title": "Building Swarm-based agents with AG2",
        "url": "https://docs.ag2.ai/docs/blog/2024-11-17-Swarm/index",
        "error": false
    },
    {
        "title": "Introducing CaptainAgent for Adaptive Team Building",
        "url": "https://docs.ag2.ai/docs/blog/2024-11-15-CaptainAgent/index",
        "error": false
    },
    {
        "title": "Unlocking the Power of Agentic Workflows at Nexla with Autogen",
        "url": "https://docs.ag2.ai/docs/blog/2024-10-23-NOVA/index",
        "error": false
    },
    {
        "title": "AgentOps, the Best Tool for AutoGen Agent Observability",
        "url": "https://docs.ag2.ai/docs/blog/2024-07-25-AgentOps/index",
        "error": false
    },
    {
        "title": "Enhanced Support for Non-OpenAI Models",
        "url": "https://docs.ag2.ai/docs/blog/2024-06-24-AltModels-Classes/index",
        "error": false
    },
    {
        "title": "AgentEval: A Developer Tool to Assess Utility of LLM-powered Applications",
        "url": "https://docs.ag2.ai/docs/blog/2024-06-21-AgentEval/index",
        "error": false
    },
    {
        "title": "Agents in AutoGen",
        "url": "https://docs.ag2.ai/docs/blog/2024-05-24-Agent/index",
        "error": false
    },
    {
        "title": "AutoDefense - Defend against jailbreak attacks with AutoGen",
        "url": "https://docs.ag2.ai/docs/blog/2024-03-11-AutoDefense/index",
        "error": false
    },
    {
        "title": "What's New in AutoGen?",
        "url": "https://docs.ag2.ai/docs/blog/2024-03-03-AutoGen-Update/index",
        "error": false
    },
    {
        "title": "StateFlow - Build State-Driven Workflows with Customized Speaker Selection in GroupChat",
        "url": "https://docs.ag2.ai/docs/blog/2024-02-29-StateFlow/index",
        "error": false
    },
    {
        "title": "FSM Group Chat -- User-specified agent transitions",
        "url": "https://docs.ag2.ai/docs/blog/2024-02-11-FSM-GroupChat/index",
        "error": false
    },
    {
        "title": "Anny: Assisting AutoGen Devs Via AutoGen",
        "url": "https://docs.ag2.ai/docs/blog/2024-02-02-AutoAnny/index",
        "error": false
    },
    {
        "title": "AutoGen with Custom Models: Empowering Users to Use Their Own Inference Mechanism",
        "url": "https://docs.ag2.ai/latest/docs/blog/2024/01/26/Custom-Models/",
        "error": false
    },
    {
        "title": "AutoGenBench -- A Tool for Measuring and Evaluating AutoGen Agents",
        "url": "https://docs.ag2.ai/docs/blog/2024-01-25-AutoGenBench/index",
        "error": false
    },
    {
        "title": "Code execution is now by default inside docker container",
        "url": "https://docs.ag2.ai/docs/blog/2024-01-23-Code-execution-in-docker/index",
        "error": false
    },
    {
        "title": "All About Agent Descriptions",
        "url": "https://docs.ag2.ai/latest/docs/blog/2023/12/29/AgentDescriptions/",
        "error": false
    },
    {
        "title": "AgentOptimizer - An Agentic Way to Train Your LLM Agent",
        "url": "https://docs.ag2.ai/docs/blog/2023-12-23-AgentOptimizer/index",
        "error": false
    },
    {
        "title": "AutoGen Studio: Interactively Explore Multi-Agent Workflows",
        "url": "https://docs.ag2.ai/docs/blog/2023-12-01-AutoGenStudio/index",
        "error": false
    },
    {
        "title": "Agent AutoBuild - Automatically Building Multi-agent Systems",
        "url": "https://docs.ag2.ai/docs/blog/2023-11-26-Agent-AutoBuild/index",
        "error": false
    },
    {
        "title": "How to Assess Utility of LLM-powered Applications?",
        "url": "https://docs.ag2.ai/docs/blog/2023-11-20-AgentEval/index",
        "error": false
    },
    {
        "title": "AutoGen Meets GPTs",
        "url": "https://docs.ag2.ai/docs/blog/2023-11-13-OAI-assistants/index",
        "error": false
    },
    {
        "title": "EcoAssistant - Using LLM Assistants More Accurately and Affordably",
        "url": "https://docs.ag2.ai/docs/blog/2023-11-09-EcoAssistant/index",
        "error": false
    },
    {
        "title": "Multimodal with GPT-4V and LLaVA",
        "url": "https://docs.ag2.ai/docs/blog/2023-11-06-LMM-Agent/index",
        "error": false
    },
    {
        "title": "AutoGen's Teachable Agents",
        "url": "https://docs.ag2.ai/docs/blog/2023-10-26-TeachableAgent/index",
        "error": false
    },
    {
        "title": "Retrieval-Augmented Generation (RAG) Applications with AutoGen",
        "url": "https://docs.ag2.ai/docs/blog/2023-10-18-RetrieveChat/index",
        "error": false
    },
    {
        "title": "Use AutoGen for Local LLMs",
        "url": "https://docs.ag2.ai/docs/blog/2023-07-14-Local-LLMs/index",
        "error": false
    },
    {
        "title": "MathChat - An Conversational Framework to Solve Math Problems",
        "url": "https://docs.ag2.ai/docs/blog/2023-06-28-MathChat/index",
        "error": false
    },
    {
        "title": "Achieve More, Pay Less - Use GPT-4 Smartly",
        "url": "https://docs.ag2.ai/docs/blog/2023-05-18-GPT-adaptive-humaneval/index",
        "error": false
    },
    {
        "title": "Does Model and Inference Parameter Matter in LLM Applications? - A Case Study for MATH",
        "url": "https://docs.ag2.ai/latest/docs/blog/2023/04/21/LLM-tuning-math",
        "error": false
    }
]
**********************************************************************

--------------------------------------------------------------------------------
INFO:httpx:HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
assistant (to user_proxy):

I have extracted all the blog posts from the specified URL. Below is the list of titles along with their corresponding links:

1. [RealtimeAgent with Gemini API](https://docs.ag2.ai/docs/blog/2025-01-29-RealtimeAgent-with-gemini/index)
2. [Tools with ChatContext Dependency Injection](https://docs.ag2.ai/docs/blog/2025-01-22-Tools-ChatContext-Dependency-Injection/index)
3. [Streaming input and output using WebSockets](https://docs.ag2.ai/docs/blog/2025-01-10-WebSockets/index)
4. [Real-Time Voice Interactions over WebRTC](https://docs.ag2.ai/docs/blog/2025-01-09-RealtimeAgent-over-WebRTC/index)
5. [Real-Time Voice Interactions with the WebSocket Audio Adapter](https://docs.ag2.ai/docs/blog/2025-01-08-RealtimeAgent-over-websocket/index)
6. [Tools Dependency Injection](https://docs.ag2.ai/docs/blog/2025-01-07-Tools-Dependency-Injection/index)
7. [Cross-Framework LLM Tool Integration with AG2](https://docs.ag2.ai/docs/blog/2024-12-20-Tools-interoperability/index)
8. [ReasoningAgent Update - Beam Search, MCTS, and LATS for LLM Reasoning](https://docs.ag2.ai/docs/blog/2024-12-20-Reasoning-Update/index)
9. [Introducing RealtimeAgent Capabilities in AG2](https://docs.ag2.ai/docs/blog/2024-12-20-RealtimeAgent/index)
10. [Knowledgeable Agents with FalkorDB Graph RAG](https://docs.ag2.ai/docs/blog/2024-12-06-FalkorDB-Structured/index)
11. [ReasoningAgent - Tree of Thoughts with Beam Search in AG2](https://docs.ag2.ai/docs/blog/2024-12-02-ReasoningAgent2/index)
12. [Agentic testing for prompt leakage security](https://docs.ag2.ai/docs/blog/2024-11-27-Prompt-Leakage-Probing/index)
13. [Building Swarm-based agents with AG2](https://docs.ag2.ai/docs/blog/2024-11-17-Swarm/index)
14. [Introducing CaptainAgent for Adaptive Team Building](https://docs.ag2.ai/docs/blog/2024-11-15-CaptainAgent/index)
15. [Unlocking the Power of Agentic Workflows at Nexla with Autogen](https://docs.ag2.ai/docs/blog/2024-10-23-NOVA/index)
16. [AgentOps, the Best Tool for AutoGen Agent Observability](https://docs.ag2.ai/docs/blog/2024-07-25-AgentOps/index)
17. [Enhanced Support for Non-OpenAI Models](https://docs.ag2.ai/docs/blog/2024-06-24-AltModels-Classes/index)
18. [AgentEval: A Developer Tool to Assess Utility of LLM-powered Applications](https://docs.ag2.ai/docs/blog/2024-06-21-AgentEval/index)
19. [Agents in AutoGen](https://docs.ag2.ai/docs/blog/2024-05-24-Agent/index)
20. [AutoDefense - Defend against jailbreak attacks with AutoGen](https://docs.ag2.ai/docs/blog/2024-03-11-AutoDefense/index)
21. [What's New in AutoGen?](https://docs.ag2.ai/docs/blog/2024-03-03-AutoGen-Update/index)
...

TERMINATE

--------------------------------------------------------------------------------

Edit this page

User Guide

​Installation

​Imports

​Agent Configuration

​Web Browsing with Crawl4AI

​Initiate Chat

Installation

Imports

Agent Configuration

Web Browsing with `Crawl4AI`

Initiate Chat