Automate Desktop Actions with AI
TuriX is an open-source computer-use agent that lets AI models click, type, and act directly on your macOS or Windows desktop.
1.0.0Add to Favorites
Why it matters
Empower your AI models to perform real-world actions on your desktop. This asset enables hands-on computer interaction for complex tasks, integrating with existing AI agents.
Outcomes
What it gets done
Execute multi-step tasks like booking flights and Ubers.
Automate content creation and document manipulation.
Control system settings and application interactions.
Integrate with AI models for advanced computer use.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/vb-turix-computer-automation-mcp | bash Capabilities
Tools your agent gets
Execute computer interaction tasks including clicking, typing, and navigation on desktop applications.
Capture current desktop screen state for visual analysis and decision-making.
Click on specified coordinates or UI elements on the desktop.
Type text input into focused applications and text fields.
Navigate to URLs or perform browser navigation actions.
Overview
TuriX Computer Automation Skill/Agent
TuriX is an open-source computer-use agent that operates a real macOS or Windows desktop for an AI model, clicking, typing, and navigating any clickable interface without needing app-specific APIs. It plugs into Claude Desktop and other agents via MCP. Use it when a task genuinely requires interacting with desktop applications - booking, document editing, multi-app workflows - rather than something achievable through a direct API call.
What it does
TuriX is a computer-use agent that gives an AI model hands-on control of a real desktop - clicking, typing, and navigating any clickable interface without needing an application-specific API. It works cross-platform on macOS and Windows, supports hot-swappable AI models via config, and integrates with Claude Desktop and other agents through MCP. The source reports internal benchmark results of a >68% pass rate on an OSWorld-style test set, with separately measured figures of over 80% on OSWorld-like macOS benchmarks and 60% on OSWorld itself.
When to use - and when NOT to
Use TuriX when you want an AI model to actually operate desktop applications on your behalf - booking travel across multiple sites, filling in documents, generating a chart and inserting it into a presentation, or toggling system settings - tasks that require real UI interaction rather than an API call. It is not the right choice for headless server automation or anything that doesn't involve an interactive desktop session, and on macOS it needs Accessibility and Safari Automation permissions granted before it can act.
Capabilities
- Computer-interaction agent benchmarked at >80% success on OSWorld-like macOS tasks and 60% on OSWorld.
- Cross-platform support for macOS and Windows (Windows via the dedicated
multi-agent-windowsbranch). - Hot-swappable AI model backend configured through
config.json. - MCP integration with Claude Desktop and other MCP-compatible agents.
- Works with any clickable interface - no application-specific APIs required.
- Qwen3.5 vision-language model support.
- Open source and free for personal and research use.
How to install
Clone the repository and set up a Python environment:
git clone https://github.com/TurixAI/TuriX-CUA
conda create -n turix_env python=3.12
conda activate turix_env
pip install -r requirements.txt
Windows users should check out the multi-agent-windows branch. Configure a task in JSON, e.g. {"agent": {"task": "open chrome, go to github, search for the repository TuriX-CUA, and star it."}}, and configure the LLM provider with provider, api_key, and base_url (e.g. https://turixapi.io/v1) fields.
Who it's for
Developers and researchers who want an open-source agent capable of operating a real desktop - clicking through apps, filling forms, and completing multi-step tasks across macOS or Windows software - rather than one limited to API-integrated tools.
Source README
TuriX lets your powerful AI models take real, hands‑on actions directly on your desktop. It ships with a state‑of‑the‑art computer‑use agent (passes > 68 % of our internal OSWorld‑style test set) yet stays 100 % open‑source and cost‑free for personal & research use.
Installation
Follow Github: https://github.com/TurixAI/TuriX-CUA
From Source Code
git clone https://github.com/TurixAI/TuriX-CUA
conda create -n turix_env python=3.12
conda activate turix_env
pip install -r requirements.txt
Windows Setup
git checkout multi-agent-windows
Configuration
Task Configuration
{
"agent": {
"task": "open chrome, go to github, search for the repository TuriX-CUA, and star it."
}
}
API Configuration
{
"llm": {
"provider": "turix",
"api_key": "YOUR_API_KEY",
"base_url": "https://turixapi.io/v1"
}
}
Capabilities
- Modern computer interaction agent with >80% success rate on OSWorld-like MAC benchmarks, 60% on OSworld.
- Cross-platform support for macOS and Windows
- Hot-swappable AI models via config.json
- MCP integration with Claude Desktop and other agents
- No application-specific APIs required-works with any clickable interface
- Qwen3.5 vision-language model support
- Open source and free for personal use and research
Usage Examples
Book a flight, hotel and uber
Search iPhone price, create Pages document, and send to contact
Generate a bar-chart in Numbers file and insert it to PowerPoint
Search video content in YouTube and like it
Open system settings, switch to Dark Mode
Notes
Requires special macOS permissions (Accessibility and Safari Automation). Windows users should use the 'windows' branch. The agent can be controlled via the MCP protocol and integrates with Claude Desktop. Performance benchmarks show superior results compared to previous open-source agents such as UI-TARS.
FAQ
Common questions
Trust
How it checks out
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.