Browser Use integrations | LangChain 简体中文

Browser Use runs browser agents and cloud browsers for AI applications:

Browser Use Cloud agents take a task in plain language, complete it in a real browser, and return the result.
Cloud browsers are remote Chrome instances you control over CDP, with stealth, proxies, live preview, and recording.
The open-source browser-use library runs browser agents on your own machine.

This page shows how to give a LangChain agent a tool that hands web tasks to a Browser Use Cloud agent.

Installation and setup

Get an API key from the Browser Use Cloud dashboard and set it in the BROWSER_USE_API_KEY environment variable:

import getpass
import os

if "BROWSER_USE_API_KEY" not in os.environ:
    os.environ["BROWSER_USE_API_KEY"] = getpass.getpass("Enter your Browser Use API key: ")

Install the Browser Use SDK:

pip

pip install -U browser-use-sdk langchain

uv add browser-use-sdk langchain

Browser agent tool

The tool below starts a Browser Use Cloud agent run, waits for it to finish, and returns the agent's answer. A finished run keeps its cloud browser alive for follow-up runs, so the tool stops that browser before returning.

from browser_use_sdk.v4 import BrowserUse
from langchain.tools import tool


@tool
def browser_use_task(task: str) -> str:
    """Complete a task in a real web browser and return the result.

    Describe the goal in plain language and include any URL to start from.
    """
    with BrowserUse() as client:
        run = client.runs.create(task)
        result = client.runs.wait_for_completion(run.id)
        # The run keeps its cloud browser alive for follow-up runs. Stop it.
        for event in client.runs.events(run.id).events:
            if event.type == "browser.ready":
                client.browsers.stop(event.data["browser_session_id"])
    if result.result is None:
        return f"Browser task ended with status {result.status}: {result.error}"
    return result.result

Call the tool directly:

browser_use_task.invoke({"task": "Open https://example.com and return the page title."})

Example Domain

Use within an agent

# pip install -qU "langchain[anthropic]" to call the model
from langchain.agents import create_agent

agent = create_agent(
    model="claude-sonnet-4-6",
    tools=[browser_use_task],
)

result = agent.invoke(
    {"messages": [{"role": "user", "content": "What is the page title of https://example.com?"}]}
)
print(result["messages"][-1].content)

The page title of [https://example.com](https://example.com) is **"Example Domain"**.

Each tool call is a billed Browser Use Cloud run. See Browser Use pricing for current rates.

Cloud browsers

To drive a browser with your own Playwright or Puppeteer code instead of an agent, create a cloud browser and connect to its CDP URL. See the Browser Use browser quickstart.

API reference

For sessions, structured output, profiles, and other options, see the Browser Use documentation.