Skip to content
Back to search

TuriX-CUA

This is the official website for TuriX Computer-use-Agent

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/turixai--turix-cua/badge.svg)](/listings/turixai--turix-cua)

HTML

<a href="/listings/turixai--turix-cua"><img src="/api/listings/turixai--turix-cua/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 374 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. <p align="center">
  2. <img src="./doc/LogoHFitted.svg" width="1600" alt="TuriX logo">
  3. </p>
  4. <h1 align="center">TuriX · Desktop Actions, Driven by AI</h1>
  5. <p align="center"><strong>Talk to your computer, watch it work.</strong></p>
  6. <p align="center">
  7. <a href="README.md">English</a> | <a href="README.zh-CN.md">中文</a>
  8. </p>
  9. ## 📞 Contact & Community
  10. Join our Discord community for support, discussions, and updates:
  11. <p align="center">
  12. <a href="https://discord.gg/yaYrNAckb5">
  13. <img src="https://img.shields.io/discord/1400749393841492020?color=7289da&label=Join%20our%20Discord&logo=discord&logoColor=white&style=for-the-badge" alt="Join our Discord">
  14. </a>
  15. </p>
  16. Or contact us with email: contact@turix.ai
  17. TuriX lets your powerful AI models take real, hands‑on actions directly on your desktop.
  18. It ships with a **state‑of‑the‑art computer‑use agent** (achieves 80% success rate on our OSWorld‑style Mac benchmark and 64.2% success rate on OSWorld) yet stays 100 % open‑source and cost‑free for personal & research use.
  19. Prefer your own model? **Change in `config.json` and go.**
  20. ## Table of Contents
  21. - [📞 Contact & Community](#-contact--community)
  22. - [🤖 OpenClaw Skill](#-openclaw-skill)
  23. - [📰 Latest News](#-latest-news)
  24. - [🖼️ Demos](#️-demos)
  25. - [✨ Key Features](#-key-features)
  26. - [📊 Model Performance](#-model-performance)
  27. - [🚀 Quick‑Start (macOS 15+)](#-quickstart-macos-15)
  28. - [1. Download the App](#1-download-the-app)
  29. - [2. Create a Python 3.12 Environment](#2-create-a-python-312-environment)
  30. - [3. Grant macOS Permissions](#3-grant-macos-permissions)
  31. - [3.1 Accessibility](#31-accessibility)
  32. - [3.2 Safari Automation](#32-safari-automation)
  33. - [4. Configure & Run](#4-configure--run)
  34. - [4.4 Skills (Optional)](#44-skills-optional)
  35. - [🤝 Contributing](#-contributing)
  36. - [🗺️ Roadmap](#️-roadmap)
  37. ---
  38. ## 🤖 OpenClaw Skill
  39. Use TuriX via OpenClaw with our published ClawHub skill:
  40. https://clawhub.ai/Tongyu-Yan/turix-cua
  41. This repo also includes local OpenClaw skill packages in `OpenCLaw_TuriX_skill/`:
  42. - macOS package in `main` (`SKILL.md` + `scripts/run_turix.sh`)
  43. - Windows package in `multi-agent-windows` (`SKILL.md` + `scripts/run_turix.ps1` + `agents/openai.yaml`)
  44. For installation and permissions, follow `OpenCLaw_TuriX_skill/README.md`.
  45. ---
  46. ## 📰 Latest News
  47. **May 11, 2026** - Now can download **TuriX** from our [official web page](https://www.ngtechai.com).
  48. **April 8, 2026** - 🚀 Introducing **TuriX 3.0.0-alpha**
  49. [Download the app](https://www.ngtechai.com/)
  50. This is our all-in-one productivity app that combines **TuriX CUA + CLI** in one workflow, and adds two new capabilities:
  51. - **TuriX-work** for everyday office execution and task orchestration
  52. - **TuriX-code** for coding, automation, and engineering tasks
  53. From writing code to handling office tasks, you can execute with CLI precision and close the loop through GUI actions in one continuous flow.
  54. **March 16, 2026** - 🐧 **Linux support is now available** on branch `multi-agent-linux`. If you want to run TuriX on Linux (for example Ubuntu), switch to that branch first:
  55. ```bash
  56. git checkout multi-agent-linux
  57. ```
  58. **March 9, 2026** - Added a new **OpenClaw Flash/Fast Mode skill for macOS** on branch `mac_legacy`. If you want to use this faster, lighter setup, switch to that branch first:
  59. ```bash
  60. git checkout mac_legacy
  61. ```
  62. **March 5, 2026** - Updated the **Windows OpenClaw local skill** on branch `multi-agent-windows` with direct dispatch, safer pre-flight checks, and the new `OpenCLaw_TuriX_skill/agents/openai.yaml`.
  63. **Earlier updates (Jan 2026 and before)** - We shipped v0.3 (DuckDuckGo, Ollama, recoverable memory compression, Skills), published the TuriX OpenClaw skill on ClawHub, upgraded the core architecture to multi-model, and rolled out major model capability improvements including Qwen3-VL support.
  64. Ready to level up? Update your `config.json` and start automating—happy hacking! 🎉
  65. *Stay tuned to our [Discord](https://discord.gg/vkEYj4EV2n) for tips, user stories, and the next big drop.*
  66. ---
  67. ## 🖼️ Demos
  68. <p align="center"><strong>TuriX App Demo</strong></p>
  69. <p align="center">
  70. <img src="./doc/app_demo.png" width="1600" alt="TuriX app demo">
  71. </p>
  72. <h3 align="center">MacOS Demo</h3>
  73. <p align="center"><strong>Book a flight, hotel and uber.</strong></p>
  74. <p align="center">
  75. <img src="./doc/booking_demo.gif" width="1600" alt="TuriX macOS demo - booking">
  76. </p>
  77. <p align="center"><strong>Search iPhone price, create Pages document, and send to contact</strong></p>
  78. <p align="center">
  79. <img src="./doc/demo1.gif" width="1600" alt="TuriX macOS demo - iPhone price search and document sharing">
  80. </p>
  81. <p align="center"><strong>Generate a bar-chart in the numbers file sent by boss in discord and insert it to the right place of my powerpoint, and reply my boss.</strong></p>
  82. <p align="center">
  83. <img src="./doc/complex_demo_mac.gif" width="1600" alt="TuriX macOS demo - excel graph to powerpoint">
  84. </p>
  85. <h3 align="center">Windows Demo</h3>
  86. <p align="center"><strong>Search video content in youtube and like it</strong></p>
  87. <p align="center">
  88. <img src="./doc/win_demo1.gif" width="1600" alt="TuriX Windows demo - video search and sharing">
  89. </p>
  90. <h3 align="center">MCP with Claude Demo</h3>
  91. <p align="center"><strong>Claude search for AI news, and call TuriX with MCP, write down the research result to a pages document and send it to contact</strong></p>
  92. <p align="center">
  93. <img src="./doc/mcp_demo1.gif" width="1600" alt="TuriX MCP demo - news search and sharing">
  94. </p>
  95. ---
  96. ## ✨ Key Features
  97. | Capability | What it means |
  98. |------------|---------------|
  99. | **SOTA default model** | Outperforms previous open‑source agents (e.g. UI‑TARS) on success rate and speed on Mac |
  100. | **No app‑specific APIs** | If a human can click it, TuriX can too—WhatsApp, Excel, Outlook, in‑house tools… |
  101. | **Hot‑swappable "brains"** | Replace the VLM policy without touching code (`config.json`) |
  102. | **MCP‑ready** | Hook up *Claude for Desktop* or **any** agent via the Model Context Protocol (MCP) |
  103. | **Skills (markdown playbooks)** | Planner selects relevant skill guides (name + description), brain uses full instructions to plan each step |
  104. ---
  105. ## 📊 Model Performance
  106. Our agent achieves state-of-the-art performance on desktop automation tasks:
  107. ### OSWorld Benchmark — 3rd Place on the Leaderboard (50 Steps)
  108. TuriX scores **64.2% (229.88 / 358)** on the full OSWorld benchmark, ranking **3rd overall** among all submitted agents. Notably, TuriX is built and optimized for **macOS**, where we achieve an **80%+ success rate** on our self-hosted OSWorld-style Mac benchmark. We used **zero Linux training data**, yet still achieve a top-3 finish on OSWorld's Linux-based environment.
  109. <p align="center">
  110. <img src="./doc/os-world.png" width="600" alt="TuriX OSWorld benchmark score — 64.2%">
  111. </p>
  112. <p align="center">
  113. <img src="./doc/performance_sum.jpg" width="1600" alt="TuriX performance">
  114. </p>
  115. For more details, check our [report](https://turix.ai/technical-report/).
  116. ## 🚀 Quick‑Start (macOS 15+)
  117. > **We never collect data**—install, grant permissions, and hack away.
  118. > **0. Windows Users**: Switch to the `multi-agent-windows` branch for Windows-specific setup and installation instructions.
  119. >
  120. > ```bash
  121. > git checkout multi-agent-windows
  122. > ```
  123. >
  124. > For the updated OpenClaw Windows local skill package, see `OpenCLaw_TuriX_skill/README.md` in that branch.
  125. >
  126. > **0. Linux Users**: Switch to the `multi-agent-linux` branch for Linux-specific setup and installation instructions.
  127. >
  128. > ```bash
  129. > git checkout multi-agent-linux
  130. > ```
  131. >
  132. > **0. Windows Legacy Users**: For the previous Windows setup, switch to the `windows_legacy` branch.
  133. >
  134. > **0. macOS Legacy Users**: For the previous single-model macOS setup, switch to the `mac_legacy` branch.
  135. ### 1. Download the App
  136. For easier usage, [download the app](https://www.ngtechai.com/)
  137. Or follow the manual setup below:
  138. ### 2. Create a Python 3.12 Environment
  139. Firstly Clone the repository and run:
  140. ```bash
  141. conda create -n turix_env python=3.12
  142. conda activate turix_env # requires conda ≥ 22.9
  143. pip install -r requirements.txt
  144. ```
  145. ### 3. Grant macOS Permissions
  146. #### 3.1 Accessibility
  147. 1. Open **System Settings ▸ Privacy & Security ▸ Accessibility**
  148. 2. Click **+**, then add **Terminal** and **Visual Studio Code** ANY IDE you use
  149. 3. If the agent still fails, also add **/usr/bin/python3**
  150. #### 3.2 Safari Automation
  151. 1. **Safari ▸ Settings ▸ Advanced** → enable **Show features for web developers**
  152. 2. In the new **Develop** menu, enable
  153. * **Allow Remote Automation**
  154. * **Allow JavaScript from Apple Events**
  155. ##### Trigger the Permission Dialogs (run once per shell)
  156. ```
  157. # macOS Terminal
  158. osascript -e 'tell application "Safari" \
  159. to do JavaScript "alert(\"Triggering accessibility request\")" in document 1'
  160. # VS Code integrated terminal (repeat to grant VS Code)
  161. osascript -e 'tell application "Safari" \
  162. to do JavaScript "alert(\"Triggering accessibility request\")" in document 1'
  163. ```
  164. > **Click "Allow" on every dialog** so the agent can drive Safari.
  165. ### 4. Configure & Run
  166. #### 4.1 Edit Task Configuration
  167. > [!IMPORTANT]
  168. > **Task Configuration is Critical**: The quality of your task instructions directly impacts success rate. Clear, specific prompts lead to better automation results.
  169. Edit task in `examples/config.json`:
  170. ```json
  171. {
  172. "agent": {
  173. "task": "open system settings, switch to Dark Mode"
  174. }
  175. }
  176. ```
  177. #### 4.2 Edit API Configuration
  178. In this main (multi-agent) branch, you need to set the brain, actor, and memory models. It only supports mac for now. If you enable planning
  179. (`agent.use_plan: true`), you also need to set the planner model.
  180. Edit API in `examples/config.json`:
  181. ```json
  182. "brain_llm": {
  183. "provider": "turix",
  184. "model_name": "turix-brain",
  185. "api_key": "YOUR_API_KEY",
  186. "base_url": "https://your-endpoint/v1"
  187. },
  188. "actor_llm": {
  189. "provider": "turix",
  190. "model_name": "turix-actor",
  191. "api_key": "YOUR_API_KEY",
  192. "base_url": "https://your-endpoint/v1"
  193. },
  194. "memory_llm": {
  195. "provider": "turix",
  196. "model_name": "turix-brain",
  197. "api_key": "YOUR_API_KEY",
  198. "base_url": "https://your-endpoint/v1"
  199. },
  200. "planner_llm": {
  201. "provider": "turix",
  202. "model_name": "turix-brain",
  203. "api_key": "YOUR_API_KEY",
  204. "base_url": "https://your-endpoint/v1"
  205. }
  206. ```
  207. For a local Ollama setup, point each role to your Ollama server:
  208. ```json
  209. "brain_llm": {
  210. "provider": "ollama",
  211. "model_name": "llama3.2-vision",
  212. "base_url": "http://localhost:11434"
  213. },
  214. "actor_llm": {
  215. "provider": "ollama",
  216. "model_name": "llama3.2-vision",
  217. "base_url": "http://localhost:11434"
  218. },
  219. "memory_llm": {
  220. "provider": "ollama",
  221. "model_name": "llama3.2-vision",
  222. "base_url": "http://localhost:11434"
  223. },
  224. "planner_llm": {
  225. "provider": "ollama",
  226. "model_name": "llama3.2-vision",
  227. "base_url": "http://localhost:11434"
  228. }
  229. ```
  230. #### 4.3 Configure Custom Models (Optional)
  231. If you want to use other models not defined by the build_llm function in the main.py, you need to first define it, then setup the config.
  232. main.py:
  233. ```
  234. if provider == "name_you_want":
  235. return ChatOpenAI(
  236. model="gpt-4.1-mini", api_key=api_key, temperature=0.3
  237. )
  238. ```
  239. Switch between ChatOpenAI, ChatGoogleGenerativeAI, ChatAnthropic, or ChatOllama base on your llm. Also change the model name.
  240. #### 4.4 Skills (Optional)
  241. Skills are lightweight markdown playbooks stored in a single folder (default: `skills/`). Each skill file starts with YAML frontmatter containing `name` and `description`, followed by the instructions. The planner only sees the name + description to select relevant skills; the brain receives the full skill content to guide step goals.
  242. Skills selection requires planning (`agent.use_plan: true`).
  243. Example skill file (`skills/github-web-actions.md`):
  244. ```md
  245. ---
  246. name: github-web-actions
  247. description: Use when navigating GitHub in a browser (searching repos, starring, etc.).
  248. ---
  249. # GitHub Web Actions
  250. - Open GitHub, use the site search, and navigate to the repo page.
  251. - If login is required, ask the user before proceeding.
  252. - Confirm the Star button state before moving on.
  253. ```
  254. Enable in `examples/config.json`:
  255. ```json
  256. {
  257. "agent": {
  258. "use_plan": true,
  259. "use_skills": true,
  260. "skills_dir": "skills",
  261. "skills_max_chars": 4000
  262. }
  263. }
  264. ```
  265. #### 4.5 Start the Agent
  266. ```bash
  267. python examples/main.py
  268. ```
  269. **Enjoy hands‑free computing 🎉**
  270. #### 4.6 Resume a Terminated Task
  271. To resume a task after an interruption, set a stable `agent_id` and enable `resume` in `examples/config.json`:
  272. ```json
  273. {
  274. "agent": {
  275. "resume": true,
  276. "agent_id": "my-task-001"
  277. }
  278. }
  279. ```
  280. Notes:
  281. - Use the same `agent_id` as the run you want to resume.
  282. - Keep the same `task` when resuming.
  283. - Resume only works if prior memory exists at `src/agent/temp_files/<agent_id>/memory.jsonl`.
  284. - To start fresh, set `resume` to `false`, change `agent_id`, or delete `src/agent/temp_files/<agent_id>`.
  285. ## 🤝 Contributing
  286. We welcome contributions! Please read our [Contributing Guide](CONTRIBUTING.MD) to get started.
  287. Quick links:
  288. - [Development Setup](CONTRIBUTING.MD#development-setup)
  289. - [Code Style Guidelines](CONTRIBUTING.MD#code-style-guidelines)
  290. - [Testing](CONTRIBUTING.MD#testing)
  291. - [Pull Request Process](CONTRIBUTING.MD#pull-request-process)
  292. For bug reports and feature requests, please [open an issue](https://github.com/TurixAI/TuriX-CUA/issues).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.