Cua GTM brain

category

market profile

• text: Computer-use agents need full desktop environments and access to existing apps, not just a shell or browser.
• text: Cua Spaces uses customer-owned or customer-rented compute, with remote access and shared control as the software layer.
• text: Alternative agent environments include E2B and Daytona, though their sandbox capabilities differ from shared desktops.

personas

• title: AI agent developer
  pains:

  • An agent needs a desktop with apps and permissions ready to use.
  • Automating native apps without disrupting the developer's own cursor is difficult.
• title: Engineering lead building an agent product
  pains:

  • Teams need to inspect and intervene in agent work.
  • Managing desktop environments across machines and cloud accounts creates operational overhead.
• title: Technical founder or individual power user
  pains:

  • Agents need access to existing signed-in apps.
  • Buying dedicated compute can add cost when spare computers already exist.

icp

• text: Small engineering teams building AI agents that interact with desktop applications are a direct fit.
• text: Teams with Mac or Linux machines or their own cloud VMs can bring existing compute; the Teams plan supports 1 to 100 seats.
• text: Individual developers using coding agents can begin locally before adopting shared team infrastructure.

message pillars

• text: Pool existing computers for agent-accessible desktops.
• text: Enable people and agents to collaborate in the same environment.
• text: Move existing app sessions into agent environments with user approval.
• text: Provide a separately usable open-source desktop driver.

objections

• objection: Will agents expose my signed-in sessions?
  response: Explain the approval flow, Touch ID for sensitive items and per-Space end-to-end sealing, then validate against the buyer's security requirements.
• objection: Does the subscription include compute?
  response: No. The customer supplies or rents the computers and pays the cloud provider separately.
• objection: Is the entire product open source?
  response: Distinguish the MIT-licensed driver from Spaces, which the pricing page describes as source-available under FSL.

competitors

• name: E2B
  why: Agent sandbox infrastructure competes for execution workloads.
• name: Daytona
  why: Developer sandboxes for agent execution are an alternative environment.
• name: Simular
  website: https://www.simular.ai/alternatives/cloud-computers-for-ai-agents
  why: Its box product offers an agent cloud desktop.

influencers

• name: Christopher Settles
  url: https://www.linkedin.com/in/christopher-settles
  why: Writes about Cua Spaces and human collaboration with agents.

keywords

• AI agent desktop sandbox
• computer use agent infrastructure
• shared desktop AI agents
• macOS VM for AI agents
• open source computer use driver
• agent cloud computer

voice

• Developer-oriented
• Direct and practical
• Collaborative

signals

• text: A team is building Claude Code, Codex or other agents that must operate native apps.
• text: Engineers are seeking shared observation and human intervention in agent desktop sessions.
• text: A team has spare Macs or its own cloud VMs and wants to pool them for agent workloads.

stack

• macOS
• Linux
• Windows preview
• QUIC
• Lume

hiring

• Founding Technical GTM Lead
• Research Intern (Summer 2026)
• Founding Engineer, Infra & Agent Systems

pricing

motion

channels

• GitHub
• developer documentation
• X
• LinkedIn
• developer events

confidence

Founders

• name: Francesco Bonacci
  title: Founder/CTO
  role: CTO
  bio: Give every agent a cloud desktop at Cua. Previously at Xbox, Microsoft AI.