Added BrowserUse Tool

.gitignore

@@ -8,4 +8,7 @@ test.py
 docs_crew/
 chroma.sqlite3
 venv
-.venv/
+.venv/
+agent_history.gif
+poetry.lock
+uv.lock

crewai_tools/tools/__init__.py

@@ -1,5 +1,6 @@
 from .ai_mind_tool.ai_mind_tool import AIMindTool
 from .brave_search_tool.brave_search_tool import BraveSearchTool
+from .browser_use_tool.browser_use_tool import BrowserUseTool
 from .browserbase_load_tool.browserbase_load_tool import BrowserbaseLoadTool
 from .code_docs_search_tool.code_docs_search_tool import CodeDocsSearchTool
 from .code_interpreter_tool.code_interpreter_tool import CodeInterpreterTool

crewai_tools/tools/browser_use_tool/README.md

@@ -0,0 +1,141 @@
+# Browser-Use Tool
+
+## Description
+
+This tool is used to give the Agent the ability to use a browser to interact with the environment.
+It is useful when the Agent needs to interact with it.
+
+The tool uses [browser-use/browser-use](https://github.com/browser-use/browser-use).
+
+## Requirements
+
+- [[email protected]](https://github.com/browser-use/browser-use/tree/0.1.16) (will be installed automatically). Later versions do not work.
+
+## Installation
+
+Install the crewai_tools package
+
+```shell
+pip install 'crewai[tools]'
+```
+
+## Examples
+
+Notes:
+
+- browser-use requires multimodal models like gpt-4o, gpt-4o-mini, etc.
+- Browser-use uses its own llm, which is different than the Agent's llm.
+The only supported llms are from langchain.
+- browser-use runs asynchronously. It creates a new event loop and runs the browser in that loop and deletes the loop after the tool is deleted. If you want to use your own event loop, you can pass it to the tool.
+- browser-use uses a Browser and a BrowserContext to run the browser. If you don't specify them, the tool will create its own Browser and BrowserContext each time the tool is called. If you want to use a consistent browser throughout the Agent's lifecycle, you should specify your own Browser and BrowserContext and manage them yourself outside the tool.
+
+### Short Examples
+
+```python
+from crewai_tools import BrowserUseTool
+from langchain_openai.chat_models import ChatOpenAI
+
+browser_use_tool = BrowserUseTool(llm=ChatOpenAI(model="gpt-4o"))
+
+Agent(
+    ...
+    tools=[browser_use_tool],
+)
+```
+
+To use a consistent browser throughout the Agent's lifecycle, you can do this
+
+```python
+import asyncio
+
+from crewai_tools import BrowserUseTool
+from langchain_openai.chat_models import ChatOpenAI
+
+browser = Browser(config=BrowserConfig(headless=False))
+
+browser_context = BrowserContext(browser=browser)
+
+browser_use_tool = BrowserUseTool(
+    llm=ChatOpenAI(model="gpt-4o"),
+    browser=browser,
+    browser_context=browser_context,
+)
+
+Agent(
+    ...
+    tools=[browser_use_tool],
+)
+
+with asyncio.new_event_loop() as loop:
+    loop: asyncio.AbstractEventLoop
+    loop.run_until_complete(browser.close())
+```
+
+### Full Working Example
+
+Found in [tests/tools/test_browser_use_tool.py](/tests/tools/test_browser_use_tool.py)
+
+```python
+import asyncio
+
+from browser_use import Browser, BrowserConfig
+from browser_use.browser.context import BrowserContext
+from crewai import Agent, Crew, Task
+
+from crewai_tools.tools.browser_use_tool import BrowserUseTool
+from langchain_openai.chat_models import ChatOpenAI
+
+
+def main():
+
+    browser = Browser(config=BrowserConfig(headless=False))
+
+    browser_context = BrowserContext(browser=browser)
+
+    browser_use_tool = BrowserUseTool(
+        llm=ChatOpenAI(model="gpt-4o"),
+        browser=browser,
+        browser_context=browser_context,
+    )
+
+    agent = Agent(
+        role="Browser Use Agent",
+        goal="Use the browser",
+        backstory=(
+            "You are the best Browser Use Agent in the world. "
+            "You have a browser that you can interact with using natural language instructions."
+        ),
+        tools=[browser_use_tool],
+        verbose=True,
+        llm="gpt-4o",
+    )
+
+    task = Task(
+        name="Navigate to webpage and summarize article",
+        description="Navigate to {webpage} and find the article about 'xAI (company)' and summarize it.",
+        expected_output="A summary of the article",
+        agent=agent,
+    )
+
+    crew = Crew(
+        tasks=[task],
+        agents=[agent],
+        verbose=True,
+    )
+
+    crew_result = crew.kickoff(
+        inputs={
+            "webpage": "https://www.wikipedia.org/",
+        }
+    )
+
+    print(crew_result.raw)
+
+    with asyncio.new_event_loop() as loop:
+        loop: asyncio.AbstractEventLoop
+        loop.run_until_complete(browser.close())
+
+
+if __name__ == "__main__":
+    main()
+```

crewai_tools/tools/browser_use_tool/__init__.py

@@ -0,0 +1,7 @@
+from .browser_use_tool import BrowserUseTool, BaseBrowserUseTool, BaseBrowserUseToolSchema
+
+__all__ = [
+    "BrowserUseTool",
+    "BaseBrowserUseTool",
+    "BaseBrowserUseToolSchema",
+]