Gå til hovedindhold
Autonomous Browser Agents & Headless FleetsOpen SourceAutonomous Web Agent

Browser-use

Autonomous open-source AI agent browser automation engine

Browser-use enables LLMs to interact directly with web browsers — clicking buttons, filling forms, and navigating complex workflows like a human.

Kom hurtigt i gang

agent.py

Kom i gang med Browser-use i dit projekt.

import asyncio
from browser_use import Agent
from langchain_anthropic import ChatAnthropic

async def main():
    agent = Agent(
        task="Go to skat.dk and find the latest tax deduction rates for commuting (befordringsfradrag).",
        llm=ChatAnthropic(model_name="claude-3-7-sonnet-20250219"),
    )
    result = await agent.run()
    print(result)

asyncio.run(main())

Om Browser-use

Browser-use is the premier open-source autonomous web browsing agent. Powered by Playwright and vision-language models, Browser-use extracts interactive elements with bounding boxes and executes multi-step web interactions autonomously, enabling agents to operate Danish services that have no machine APIs.

DK

Danske data- og produkt-guides

Hvordan udviklere og agenter bruger Browser-use til at tilgå dansk digital infrastruktur.

Automate Boligsiden Property Search & Filtering

Boligsiden Real Estate

Instruct an agent to open Boligsiden, set regional filters (e.g. København K, max 6M DKK), and compile a spreadsheet of newly listed apartments.

from browser_use import Agent from langchain_openai import ChatOpenAI agent = Agent( task="Go to boligsiden.dk, search for apartments in 2100 København Ø under 5.000.000 DKK, and save the top 3 links.", llm=ChatOpenAI(model="gpt-4o"), ) await agent.run()

Nøglefunktioner

Vision + DOM tree interactive element extraction
Supports persistent browser sessions with saved cookies
Multi-tab management and automated file downloads
Integrates with LangChain, LangGraph, and Claude 3.7 / GPT-4o
Human-in-the-loop takeover mode for MitID or 2FA checkpoints