<div align="center">
<h1>🪟 Windows-Use</h1>
<a href="https://pepy.tech/project/windows-use">
<img src="https://static.pepy.tech/badge/windows-use" alt="PyPI Downloads">
</a>
<a href="https://github.com/CursorTouch/windows-use/blob/main/LICENSE">
<img src="https://img.shields.io/badge/license-MIT-green" alt="License">
</a>
<img src="https://img.shields.io/badge/python-3.12%2B-blue" alt="Python">
<img src="https://img.shields.io/badge/platform-Windows%207–11-blue" alt="Platform: Windows 7 to 11">
<br>
<a href="https://x.com/CursorTouch">
<img src="https://img.shields.io/badge/follow-%40CursorTouch-1DA1F2?logo=twitter&style=flat" alt="Follow on Twitter">
</a>
<a href="https://discord.com/invite/Aue9Yj2VzS">
<img src="https://img.shields.io/badge/Join%20on-Discord-5865F2?logo=discord&logoColor=white&style=flat" alt="Join us on Discord">
</a>
</div>
<br>
**Windows-Use** is a powerful automation agent that interact directly with the Windows at GUI layer. It bridges the gap between AI Agents and the Windows OS to perform tasks such as opening apps, clicking buttons, typing, executing shell commands, and capturing UI state all without relying on traditional computer vision models. Enabling any LLM to perform computer automation instead of relying on specific models for it.
## 🛠️Installation Guide
### **Prerequisites**
- Python 3.12 or higher
- [UV](https://github.com/astral-sh/uv) (or `pip`)
- Windows 7 or 8 or 10 or 11
### **Installation Steps**
**Install using `uv`:**
```bash
uv pip install windows-use
````
Or with pip:
```bash
pip install windows-use
```
## ⚙️Basic Usage
```python
# main.py
from langchain_google_genai import ChatGoogleGenerativeAI
from windows_use.agent import Agent
from dotenv import load_dotenv
load_dotenv()
llm=ChatGoogleGenerativeAI(model='gemini-2.0-flash')
agent = Agent(llm=llm,browser='chrome',use_vision=True)
query=input("Enter your query: ")
agent_result=agent.invoke(query=query)
print(agent_result.content)
```
## 🤖 Run Agent
You can use the following to run from a script:
```bash
python main.py
Enter your query: <YOUR TASK>
```
---
## 🎥 Demos
**PROMPT:** Write a short note about LLMs and save to the desktop
<https://github.com/user-attachments/assets/0faa5179-73c1-4547-b9e6-2875496b12a0>
**PROMPT:** Change from Dark mode to Light mode
<https://github.com/user-attachments/assets/47bdd166-1261-4155-8890-1b2189c0a3fd>
## 🪴Grounding
## Vision
Talk to your computer. Watch it get things done.
## Roadmap
### 🤖 Agent Intelligence
* [ ] **Integrate memory** : allow the agent to remember past interactions made by the user.
* [ ] **Optimize token usage** : implement strategies like Ally Tree compression and prompt engineering to reduce overhead.
* [ ] **Simulate advanced human-like input** : enable accurate and naturalistic mouse & keyboard interactions across apps.
* [ ] **Support for local LLMs** : local models with near-parity performance to cloud-based APIs (e.g., Mistral, LLaMA, etc.).
* [ ] **Improve reasoning and planning** : enhance the agent's ability to break down and sequence complex tasks.
### 🌳 Ally Tree Optimization
* [ ] **Improve UI element detection** : automatically identify and prioritize essential, interactive components on screen.
* [ ] **Compress Ally Tree intelligently** : reduce complexity by pruning irrelevant branches.
* [ ] **Context-aware prioritization** : rank UI elements based on relevance to the task at hand.
### 💡 User Experience
* [ ] **Reduce latency** : optimize to improve response time between GUI interaction.
* [ ] **Polish command interface** : make it easier to write, speak, or type commands through a simplified UX layer.
* [ ] **Better error handling & recovery** : ensure graceful handling of edge cases and unclear instructions.
### 🧪 Evaluation
* [ ] **LLM evaluation benchmarks** — track performance across different models and benchmarks.
## ⚠️ Caution
Agent interacts directly with your Windows OS at GUI layer to perform actions. While the agent is designed to act intelligently and safely, it can make mistakes that might bring undesired system behaviour or cause unintended changes. Try to run the agent in a sandbox envirnoment.
## 🪪 License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## 🤝 Contributing
Contributions are welcome! Please check the [CONTRIBUTING](CONTRIBUTING) file for setup and development workflow.
Made with ❤️ by [Jeomon George](https://github.com/Jeomon)
---
## Citation
```bibtex
@software{
author = {George, Jeomon},
title = {Windows-Use: Enable AI to control Windows OS},
year = {2025},
publisher = {GitHub},
url={https://github.com/CursorTouch/Windows-Use}
}
```
Raw data
{
"_id": null,
"home_page": null,
"name": "windows-use",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.13",
"maintainer_email": null,
"keywords": "agent, ai, ai agent, automation, desktop, windows",
"author": null,
"author_email": "Jeomon George <jeogeoalukka@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/18/aa/a292322d8baf1bc183b90841fef1b5c1270b67bdfd2ebebe9f8edbed07b3/windows_use-0.5.0.tar.gz",
"platform": null,
"description": "<div align=\"center\">\n\n <h1>\ud83e\ude9f Windows-Use</h1>\n <a href=\"https://pepy.tech/project/windows-use\">\n <img src=\"https://static.pepy.tech/badge/windows-use\" alt=\"PyPI Downloads\">\n </a>\n <a href=\"https://github.com/CursorTouch/windows-use/blob/main/LICENSE\">\n <img src=\"https://img.shields.io/badge/license-MIT-green\" alt=\"License\">\n </a>\n <img src=\"https://img.shields.io/badge/python-3.12%2B-blue\" alt=\"Python\">\n <img src=\"https://img.shields.io/badge/platform-Windows%207\u201311-blue\" alt=\"Platform: Windows 7 to 11\">\n <br>\n\n <a href=\"https://x.com/CursorTouch\">\n <img src=\"https://img.shields.io/badge/follow-%40CursorTouch-1DA1F2?logo=twitter&style=flat\" alt=\"Follow on Twitter\">\n </a>\n <a href=\"https://discord.com/invite/Aue9Yj2VzS\">\n <img src=\"https://img.shields.io/badge/Join%20on-Discord-5865F2?logo=discord&logoColor=white&style=flat\" alt=\"Join us on Discord\">\n </a>\n\n</div>\n\n<br>\n\n**Windows-Use** is a powerful automation agent that interact directly with the Windows at GUI layer. It bridges the gap between AI Agents and the Windows OS to perform tasks such as opening apps, clicking buttons, typing, executing shell commands, and capturing UI state all without relying on traditional computer vision models. Enabling any LLM to perform computer automation instead of relying on specific models for it.\n\n## \ud83d\udee0\ufe0fInstallation Guide\n\n### **Prerequisites**\n\n- Python 3.12 or higher\n- [UV](https://github.com/astral-sh/uv) (or `pip`)\n- Windows 7 or 8 or 10 or 11\n\n### **Installation Steps**\n\n**Install using `uv`:**\n\n```bash\nuv pip install windows-use\n````\n\nOr with pip:\n\n```bash\npip install windows-use\n```\n\n## \u2699\ufe0fBasic Usage\n\n```python\n# main.py\nfrom langchain_google_genai import ChatGoogleGenerativeAI\nfrom windows_use.agent import Agent\nfrom dotenv import load_dotenv\n\nload_dotenv()\n\nllm=ChatGoogleGenerativeAI(model='gemini-2.0-flash')\nagent = Agent(llm=llm,browser='chrome',use_vision=True)\nquery=input(\"Enter your query: \")\nagent_result=agent.invoke(query=query)\nprint(agent_result.content)\n```\n\n## \ud83e\udd16 Run Agent\n\nYou can use the following to run from a script:\n\n```bash\npython main.py\nEnter your query: <YOUR TASK>\n```\n\n---\n\n## \ud83c\udfa5 Demos\n\n**PROMPT:** Write a short note about LLMs and save to the desktop\n\n<https://github.com/user-attachments/assets/0faa5179-73c1-4547-b9e6-2875496b12a0>\n\n**PROMPT:** Change from Dark mode to Light mode\n\n<https://github.com/user-attachments/assets/47bdd166-1261-4155-8890-1b2189c0a3fd>\n\n## \ud83e\udeb4Grounding\n\n\n\n\n\n\n## Vision\n\nTalk to your computer. Watch it get things done.\n\n## Roadmap\n\n### \ud83e\udd16 Agent Intelligence\n\n* [ ] **Integrate memory** : allow the agent to remember past interactions made by the user.\n* [ ] **Optimize token usage** : implement strategies like Ally Tree compression and prompt engineering to reduce overhead.\n* [ ] **Simulate advanced human-like input** : enable accurate and naturalistic mouse & keyboard interactions across apps.\n* [ ] **Support for local LLMs** : local models with near-parity performance to cloud-based APIs (e.g., Mistral, LLaMA, etc.).\n* [ ] **Improve reasoning and planning** : enhance the agent's ability to break down and sequence complex tasks.\n\n### \ud83c\udf33 Ally Tree Optimization\n\n* [ ] **Improve UI element detection** : automatically identify and prioritize essential, interactive components on screen.\n* [ ] **Compress Ally Tree intelligently** : reduce complexity by pruning irrelevant branches.\n* [ ] **Context-aware prioritization** : rank UI elements based on relevance to the task at hand.\n\n### \ud83d\udca1 User Experience\n\n* [ ] **Reduce latency** : optimize to improve response time between GUI interaction.\n* [ ] **Polish command interface** : make it easier to write, speak, or type commands through a simplified UX layer.\n* [ ] **Better error handling & recovery** : ensure graceful handling of edge cases and unclear instructions.\n\n### \ud83e\uddea Evaluation\n\n* [ ] **LLM evaluation benchmarks** \u2014 track performance across different models and benchmarks.\n\n## \u26a0\ufe0f Caution\n\nAgent interacts directly with your Windows OS at GUI layer to perform actions. While the agent is designed to act intelligently and safely, it can make mistakes that might bring undesired system behaviour or cause unintended changes. Try to run the agent in a sandbox envirnoment.\n\n## \ud83e\udeaa License\n\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n## \ud83e\udd1d Contributing\n\nContributions are welcome! Please check the [CONTRIBUTING](CONTRIBUTING) file for setup and development workflow.\n\nMade with \u2764\ufe0f by [Jeomon George](https://github.com/Jeomon)\n\n---\n\n## Citation\n\n```bibtex\n@software{\n author = {George, Jeomon},\n title = {Windows-Use: Enable AI to control Windows OS},\n year = {2025},\n publisher = {GitHub},\n url={https://github.com/CursorTouch/Windows-Use}\n}\n```\n",
"bugtrack_url": null,
"license": null,
"summary": "An AI Agent that interacts with Windows OS at GUI level.",
"version": "0.5.0",
"project_urls": {
"homepage": "https://github.com/CursorTouch"
},
"split_keywords": [
"agent",
" ai",
" ai agent",
" automation",
" desktop",
" windows"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "2144808ee9391fc90901dc37abc37196459e9fba3d10e725fe256c2715be2626",
"md5": "e929553e95d672adb97b1cc4f5ee20ba",
"sha256": "28d90fef836644310b1c5663946ebded8126cea5bcdd108da52c33de1303e2e7"
},
"downloads": -1,
"filename": "windows_use-0.5.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "e929553e95d672adb97b1cc4f5ee20ba",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.13",
"size": 29120,
"upload_time": "2025-07-21T01:39:25",
"upload_time_iso_8601": "2025-07-21T01:39:25.545799Z",
"url": "https://files.pythonhosted.org/packages/21/44/808ee9391fc90901dc37abc37196459e9fba3d10e725fe256c2715be2626/windows_use-0.5.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "18aaa292322d8baf1bc183b90841fef1b5c1270b67bdfd2ebebe9f8edbed07b3",
"md5": "2ae4fa1672ae68ceef086260f866b3ac",
"sha256": "81d14acad5e80ff04df51a2aa13cddb2211ddf0e6799a9bcbae7ca08050f52c5"
},
"downloads": -1,
"filename": "windows_use-0.5.0.tar.gz",
"has_sig": false,
"md5_digest": "2ae4fa1672ae68ceef086260f866b3ac",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.13",
"size": 22765,
"upload_time": "2025-07-21T01:39:27",
"upload_time_iso_8601": "2025-07-21T01:39:27.762537Z",
"url": "https://files.pythonhosted.org/packages/18/aa/a292322d8baf1bc183b90841fef1b5c1270b67bdfd2ebebe9f8edbed07b3/windows_use-0.5.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-07-21 01:39:27",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "windows-use"
}
JSON