Run AI agents that see, reason, and act on any desktop
Free plan available, paid tiers aboveVisit WebsiteFreemiumVisit Website
Tracked since2026
The Bottom Line
- Price
- Free plan available, paid tiers above. Compare plans
Key facts
- Open-source infrastructure for computer-use AI agents, providing full desktop environments on Linux, Windows, macOS, and Android.
- Includes Lume for local macOS VMs on Apple Silicon and Cua Cloud for scalable cloud sandboxes.
- Backed by Y Combinator, with tools like Agent SDK, CLI, VLM Router, Playground, and Cua-Bench for benchmarking.
Pros
- Open-source and self-hostable, giving full control over infrastructure
- Supports all major desktop operating systems plus Android
- Built by researchers with deep expertise in computer-use agents (Windows Agent Arena)
Cons
- Cloud service requires booking a call for access, not self-serve
- Primarily designed for developers and researchers, not end-users
What is Cua?
Cua provides open-source infrastructure for running AI agents that can see a screen, reason about it, and act. It gives agents full desktop environments on Linux, Windows, macOS, and Android, enabling them to interact with software just like a human would. The platform includes Lume for running native macOS VMs locally on Apple Silicon, and Cua Cloud for scalable cloud sandboxes across all supported operating systems.
Built by a team that pioneered Windows Agent Arena at Microsoft, Cua is designed for developers and researchers building computer-use agents. It offers a self-hostable Agent SDK, a CLI, a VLM Router, a Playground, and Cua-Bench for benchmarking. The open-source components (Lume, Agent SDK, Cua-Bench) are MIT-licensed, while Cua Cloud provides managed desktop fleets for production use.
Key Features
- Multi-OS desktop environments for agents (Linux, Windows, macOS, Android)
- Lume: open-source CLI for running native macOS VMs on Apple Silicon
- Cua Cloud: managed cloud sandboxes for scaling agent fleets
- Agent SDK for building and orchestrating computer-use agents
- Cua-Bench: benchmark suite for evaluating computer-use agents on desktop and mobile tasks
- VLM Router and Playground for vision-language model integration
- CuaBot: multi-player computer-use for coding agents
- Self-hostable, MIT-licensed open-source components
Pricing Plans
Pricing checked Oct 1, 2026
| Plan | Price | Details |
|---|---|---|
| Local | Free |
+1 moreShow less
|
| Dedicated fleets | Contact us |
+1 moreShow less
|
Is Cua worth the price?
Fair value
Cua's pricing is fair for its niche, with a generous free Local tier that lets developers test GUI Docker, QEMU, and Apple VZ sandboxes at no cost.
The Dedicated fleets tier is opaque with 'Contact us' pricing, which may hide high costs for hosted, BYOC, or on-prem setups, making it best for teams needing scalable, secure computer-use agent infrastructure.
Hidden Costs & Gotchas
Potential overage fees for extra concurrency
Reviews
$99Free with your review
Write a reviewReview Cua, get a free AI guide
Share your experience and we will send you Improve Your Thinking Patterns Using ChatGPT, free.
Best Cua Alternatives
Top alternatives based on features, pricing, and user needs.
Still deciding?
Most buyers shortlist 2 or 3 tools before committing. Pull a side-by-side comparison or browse the full alternatives shortlist below.
Explore More
Cua FAQ
How does Cua help developers build agents that can control desktop software?
Cua provides open-source infrastructure for running AI agents that see a screen, reason about it, and act, giving them full desktop environments on Linux, Windows, macOS, and Android. Developers can use the self-hostable Agent SDK, CLI, VLM Router, Playground, and Cua-Bench to build and benchmark computer-use agents that interact with software like a human would.
Which teams benefit most from using Cua?
Cua is designed for developers and researchers building computer-use agents, particularly those who need full control over their infrastructure. Teams working on AI agent research or automation across multiple desktop operating systems will find its open-source, self-hostable components and managed cloud sandboxes most useful.
How does Cua compare to Devin for building software agents?
Unlike Devin, which focuses on autonomous software development, Cua provides open-source infrastructure for running agents that can see, reason, and act on any desktop environment across Linux, Windows, macOS, and Android. Cua is built by the team that pioneered Windows Agent Arena at Microsoft and offers self-hostable components like the Agent SDK and Cua-Bench for benchmarking.
What are the main limitations or trade-offs of using Cua?
Cua's cloud service requires booking a call for access and is not self-serve, which may slow down onboarding for some teams. Additionally, the platform is primarily designed for developers and researchers, not end-users, so non-technical users may find it less accessible.
How is Cua priced?
Cua is available on a free tier, with paid plans for more usage and features. The open-source components (Lume, Agent SDK, Cua-Bench) are MIT-licensed, while Cua Cloud provides managed desktop fleets for production use.
Can Cua run agents on both local machines and cloud environments?
Yes, Cua supports both local and cloud environments: Lume runs native macOS VMs locally on Apple Silicon, and Cua Cloud provides scalable cloud sandboxes across Linux, Windows, macOS, and Android. The platform also includes a self-hostable Agent SDK for custom deployments.
Does Cua include tools for benchmarking agent performance?
Yes, Cua includes Cua-Bench, a benchmarking tool for evaluating computer-use agents. This is part of the open-source, MIT-licensed components alongside the Agent SDK and Lume.
What kind of integrations does Cua support for agent development?
Cua offers a VLM Router, a Playground, and a CLI as part of its developer tooling, along with the Agent SDK for building custom agents. It provides full desktop environments on all major operating systems, enabling agents to interact with any software installed on those systems.
Source: trycua.com