API Reference

This section documents all public classes, methods, and functions provided by the CUA SDK.

ComputerAgent

Class: ComputerAgent

The main interface for creating and controlling agents.

Constructor

ComputerAgent(name: str = "Merlin", mode: str = "interactive", config: dict = None)

• name (str): Name of the agent (default: "Merlin")
• mode (str): Operation mode, e.g., "interactive" or "batch"
• config (dict): Optional configuration dictionary

Methods

run_task(task: str) -> Any

Runs a task using the agent.

• task (str): Description of the task to perform (e.g., "Open Notepad and type 'Hello'")
• Returns: Result of the task (type may vary)

agent = ComputerAgent()
result = agent.run_task("Open Notepad and type 'Hello'")

stop() -> None

Stops the agent and cleans up resources.

agent.stop()

Docker Utilities

If using Docker, the following scripts/utilities are available:

• Dockerfile: Builds the agent image.
• cua_docker/cua_docker_loop.py: Main entrypoint for running agents in Docker.
• cua_docker/config.py: Handles Docker-specific configuration.

Configuration Utilities

• load_dotenv(): Loads environment variables from .env.
• os.getenv("VAR_NAME"): Accesses environment variables in code.

Additional Functions

Document any other public functions or utilities here as your SDK evolves.

For more usage examples, see Usage Examples.