Cua
Docker for Computer-use Agents
📚 TL;DR
c/ua is an open-source framework that enables AI agents to control full operating systems within high-performance, lightweight virtual containers. It delivers up to 97% native CPU speed on Apple Silicon and works with any language models: https://github.com/trycua/cua.
⚒️ The Problem
For Computer-Use AI agents to be genuinely useful, they must interact with your system's native applications. But giving full access to your host device is risky - and not ideal. What if the agent's process gets compromised, or the LLM hallucinates and leaks your data? And practically speaking, do you really want to give up control of your entire machine just so the agent can do its job?
In short: How can AI agents interact with operating systems, desktop applications, and browsers without jeopardizing security or sacrificing performance?
⭐ Solution: An end-to-end framework
We combined Apple's Virtualization.Framework with an agent-friendly interface to create:
- Lightweight VMs: Run at 97% of native CPU speed on Apple Silicon, giving you near-native performance in a secure, virtualized space.
- Sandboxed Execution: AI agents operate in tightly controlled environments, allowing secure interaction with desktop applications - without risking your host system.
- LLM-Compatible Runtimes: Works out-of-the-box with popular LLM providers like OpenAI, Anthropic, and open-source tools like Ollama, LM Studio, and OpenRouter.
And we now support MCP! You can use Claude Desktop or Cursor, as your personal assistant to navigate even notoriously challenging apps like Tableau. Watch the demo to see it in action!
⚙️ How It Works
c/ua brings together three powerful components into one unified framework stack:
1. Lume - Our high-performance Virtualization layer leverages Apple's Virtualization.Framework to spin up macOS or Linux environments.
2. Computer-Use Interface (CUI) - This layer lets any process or AI agent - no matter which framework it's built on (like AutoGen, LangGraph, etc.) - see the screen and interact with Lume VMs just like a human would. It supports actions such as clicking, typing, scrolling, and extracting the visual accessibility tree.
3. Computer-Use Agent (CUA) - Built on top of CUI, this general agent framework runs intelligent workflows within these VMs. It's compatible with popular LLM providers like OpenAI, Anthropic, as well as open-source tools like Ollama, LM Studio, and OpenRouter.
An AI agent built with c/ua operates within a virtual environment to:
- Navigate and interact with any application's interface
- Read screen content and perform keyboard/mouse actions
- Switch between applications and self-debug when needed
- Operate in a secure sandbox with controlled file access
All of this occurs in a fully isolated environment, ensuring your host system, files, and sensitive data remain completely secure - while you continue using your device without interruption.
🖥️ What Can You Build with c/ua?
c/ua lets AI agents use real desktop apps - just like a human. For example:
- CAD Copilot: Modify 3D models in tools like Fusion 360 or FreeCAD.
- Photoshop Assistant: Crop, layer, and export assets via prompt.
- Operator for Tableau: Navigate dashboards and filters with natural language.
- Legacy App Automator: Control apps and websites with no APIs through UI actions.
🤝 Our Asks:
If you're done burning money on fragile browser automations or slow, over-engineered computer-use agents - it's time to try c/ua:
- Book a call: https://cal.com/francesco-bonacci/cua-demo
- Email us: founders@trycua.com
Also don’t forget to:
1. Leave us a star, watch and fork on GitHub: https://github.com/trycua/cua
2. Join our Discord https://discord.com/invite/mVnXXpdE85 - we're building this together
