Enable AI agents to control your desktop (Windows, MacOS, Linux), mobile (Android, iOS) and HMI devices
Join the AskUI Discord.
Traditional UI automation is fragile. Every time a button moves, a label changes, or a layout shifts, your scripts break. You're stuck maintaining brittle selectors, writing conditional logic for edge cases, and constantly updating tests.
AskUI Agents solve this by combining two powerful approaches:
- Vision-based automation - Find UI elements by what they look like or say, not by brittle XPath or CSS selectors
- AI-powered agents - Give high-level instructions and let AI figure out the steps AskUI Python SDK is a powerful automation framework that enables you and AI agents to control your desktop, mobile, and HMI devices and automate tasks. With support for multiple AI models, multi-platform compatibility, and enterprise-ready features,
Whether you're automating desktop apps, testing mobile applications, or building RPA workflows, AskUI adapts to UI changes automatically—saving you hours of maintenance work.
- Multi-platform - Works on Windows, Linux, MacOS, and Android
- Two modes - Single-step UI commands or agentic intent-based instructions
- Vision-first - Find elements by text, images, or natural language descriptions
- Model flexibility - Anthropic Claude, Google Gemini, AskUI models, or bring your own
- Extensible - Add custom tools and capabilities via Model Context Protocol (MCP)
- Caching - Save expensive calls to AI APIs by using them only when really necessary
Run the script with python <file path>, e.g python test.py to see if it works.
In order to let AI agents control your devices, you need to be able to connect to an AI model (provider). We host some models ourselves and support several other ones, e.g. Anthropic, OpenRouter, Hugging Face, etc. out of the box. If you want to use a model provider or model that is not supported, you can easily plugin your own (see Custom Models).
For this example, we will us AskUI as the model provider to easily get started.
Sign up at hub.askui.com to:
- Activate your free trial by signing up (no credit card required)
- Get your workspace ID and access token
Linux & MacOS
export ASKUI_WORKSPACE_ID=<your-workspace-id-here>
export ASKUI_TOKEN=<your-token-here>Windows PowerShell
$env:ASKUI_WORKSPACE_ID="<your-workspace-id-here>"
$env:ASKUI_TOKEN="<your-token-here>"from askui import ComputerAgent
with ComputerAgent() as agent:
# Complex multi-step instruction
agent.act(
"Open a browser, navigate to GitHub, search for 'askui vision-agent', "
"and star the repository"
)
# Extract information from the screen
balance = agent.get("What is my current account balance?")
print(f"Balance: {balance}")
# Combine both approaches
agent.act("Find the login form")
agent.type("user@example.com")
agent.click("Next")Add new capabilities to your agents:
from askui import ComputerAgent
from askui.tools.store.computer import ComputerSaveScreenshotTool
from askui.tools.store.universal import PrintToConsoleTool
with ComputerAgent() as agent:
agent.act(
"Take a screenshot of the current screen and save it, then confirm",
tools=[
ComputerSaveScreenshotTool(base_dir="./screenshots"),
PrintToConsoleTool()
]
)Ready to build your first agent? Check out our documentation:
- Start Here - Overview and core concepts
- Setup - Installation and configuration
- Using Agents - Using the AskUI ComputerAgent and AndroidAgent
- System Prompts - How to write effective instructions
- Using Models - Using different models as backbone for act, get, and locate
- BYOM - use your own model cloud by plugging in your own model provider
- Caching - Optimize performance and costs
- Tools - Extend agent capabilities
- Reporting - Obtain agent logs as execution reports and summaries as test reports
- Observability - Monitor and debug agents
- Extracting Data - Extracting structured data from screenshots and files
Official documentation: docs.askui.com
pip install askui[all]Requires Python >=3.10, <3.14
You'll also need to install AskUI Agent OS for device control. See Setup Guide for detailed instructions.
This project is licensed under the MIT License - see the LICENSE file for details.