c/ua

c/ua (Computer-Use AI Agents) is a comprehensive, developer-focused platform designed to streamline the creation, deployment, and management of AI agents that can interact with computer operating systems. Backed by Y Combinator, c/ua addresses the critical need for secure, isolated, and scalable environments for agentic workflows, effectively removing the infrastructure headaches for developers.

The platform's core offering is its cloud-based containers, which provide sandboxed environments on macOS, Linux, and Windows. This allows AI agents to perform complex tasks—from robotic process automation (RPA) to software testing—without compromising the host system's security. c/ua is built with an open-source core, fostering a community of developers and ensuring transparency and flexibility.

How to use c/ua

Getting started with c/ua involves a few straightforward steps, catering to both local development and cloud deployment:

1. Choose Your Environment: Start with the free Open Source plan to run agents locally on your machine. For scalable, managed solutions, opt for the Pro plan to access cloud containers.
2. Integrate the SDK: Use the `cua-agent` SDK, a powerful open-source framework, to build your agentic workflows in Python. The SDK simplifies the process of defining tasks and interacting with the environment.
3. Configure Your LLM: c/ua is model-agnostic. You can easily configure your agent to use your preferred Large Language Model (LLM) from providers like OpenAI, Anthropic, ByteDance, OpenRouter, and more.
4. Deploy and Run: Execute your agent locally for testing and development. When ready, deploy it to a c/ua cloud container with a simple command. The platform handles all the complexity of provisioning and securing the environment.
5. Interact and Control: Use the Multi-modal Control Protocol (MCP) to control your computer through natural language conversations with clients like Cursor or Claude Desktop. For fine-grained programmatic control, use the `Computer` library, which offers PyAutoGUI-compatible primitives like `click`, `scroll`, and `type`.
6. Generate Training Data: Leverage the Gradio UI to record your own computer interactions, creating high-quality, human-like trajectories that can be used to train or fine-tune computer-use AI models.

Core Features of c/ua

• Secure Cloud Containers: Deploy agents in isolated, secure virtual environments across macOS, Linux, and Windows, eliminating infrastructure management.
• Open-Source Agent Framework: A flexible and powerful framework (`cua-agent`) for building agentic RPA workflows.
• Multi-LLM Provider Support: Freedom to integrate with a wide range of LLMs, including those from OpenAI, Anthropic, and OpenRouter.
• Cross-Platform Automation: Design and execute complex workflows that operate seamlessly across different operating systems.
• Natural Language Control: The Multi-modal Control Protocol (MCP) enables users to command agents using conversational language through integrated desktop clients.
• Programmatic Computer Control: A PyAutoGUI-compatible API for precise, code-based control over mouse, keyboard, and screen actions.
• Human Trajectory Generation: A dedicated tool to create high-quality datasets for training computer-use models by recording human interactions.
• Advanced VM Management: Includes tools like Lume (CLI for Apple Silicon VMs) and Lumier (Docker-based macOS VMs) for developers who need more control over their virtual environments.

Use Cases for c/ua

c/ua is versatile and can be applied to a wide range of automation and AI development scenarios:

• Intelligent Robotic Process Automation (RPA): Automate complex business processes, data entry, and report generation using AI agents that can understand context and adapt to changes.
• Automated Software Testing: Deploy agents to perform comprehensive UI/UX testing, simulating human interaction to identify bugs and usability issues.
• Autonomous Web Scraping and Data Aggregation: Create sophisticated agents that can navigate complex websites, log into accounts, and extract structured data.
• AI Model Development and Training: Generate rich, high-quality datasets of human-computer interactions to train vision and control models.
• Personal AI Assistants: Build customized agents that can manage your digital tasks, organize files, and control applications based on natural language commands.

Advantages of c/ua

c/ua offers significant advantages for developers and teams building AI agents:

• Infrastructure Abstraction: Frees developers from the complexities of setting up, securing, and scaling virtual environments.
• Enhanced Security: The sandboxed container approach ensures that agent actions are confined, protecting the underlying system.
• Flexibility and Extensibility: The open-source core and support for multiple LLMs and platforms prevent vendor lock-in and encourage customization.
• Developer-First Experience: Designed with a comprehensive SDK, clear documentation, and powerful CLI tools to streamline the development lifecycle.
• Transparent and Scalable Pricing: A clear, usage-based pricing model that scales with your needs, from individual hobbyists to large enterprises.

Pricing and Plans

c/ua offers a crystal-clear pricing structure tailored to different user needs:

• Open Source (Free): Ideal for hackers, hobbyists, and FOSS projects. This plan includes the MIT Licensed Core, unlimited local agents, and access to community support via Discord and documentation.
• Pro Plan ($75 for 5,100 credits): Designed for teams that require hosted agents without infrastructure overhead. This plan provides access to all public cloud regions (macOS, Linux, Windows). Usage is metered in universal credits (5,100 credits equate to approximately 425 hours of typical runtime). It also includes email and Slack support.
• Enterprise Plan (Custom Pricing): For organizations with large-scale requirements. This includes everything in the Pro plan, plus 24/7 support, and compliance reports like HIPAA and SOC Type 1/2. Contact sales for a quote.