cua

View on GitHub

Scale computer-use 2.0 with open-source drivers, cross-OS fleets, and benchmarks for training, evaluation, and data generation.

★ 27K1,882 forksHTMLMITtrycua

Cua is an open-source computer-use agent stack: a cross-OS driver to control native apps, isolated cloud desktop fleets, local macOS/Linux VMs via Lume, CUA-S1 decision models, and Cua Bench for evaluating agents and exporting trajectories.

Use Cases

Give LLM agents a real computer to operateAutomate macOS/Windows/Linux desktop appsProvision isolated cloud desktop fleets for agentsBenchmark and evaluate computer-use agentsGenerate trajectory data for training computer-use modelsRun local macOS/Linux VMs on Apple SiliconConnect agents to native apps via MCP or CLISandboxed browser and GUI task executionHuman-in-the-loop agent evaluation

Built With

Language: HTML
Frameworks: Apple Virtualization.Framework · MCP · FastMCP · pydantic · Playwright · OpenAI SDK · Anthropic SDK · Docker · Kasm · pytest · uv

cua

Use Cases

Built With

Tags