Daytona expands into agent desktops

This funding marks Daytona’s move from selling a single tool for running agent-written code into selling the full machine an agent works on. A sandbox lets an agent execute a script safely. A broader computer platform lets it keep files around, pause and resume work, open a desktop, use Windows or GPU environments, and handle longer jobs like testing apps, training workflows, and enterprise automation.1

• The product expansion is concrete. Daytona now offers persistent storage through shared volumes, paused virtual machines, desktop environments, network controls, and multiple operating systems. That turns a short lived code runner into rented infrastructure for agents that need memory, files, and a working desktop over time.1
• This follows a pattern seen elsewhere in AI developer infrastructure. Vercel expanded from deployment into workflows, sandboxes, and backend services as agent built apps needed long running jobs, state, and safe execution. The market is moving from code generation alone toward owning the runtime where agents actually do work.2 4
• Daytona’s founders came from browser based development environments, so the jump into agent desktops is adjacent to their original skill set. Replit showed a similar arc from browser IDE to AI agent and hosting stack, where the value shifted from editing code to generating, running, and publishing complete applications.1 3

The next step is for agent infrastructure to look less like a developer utility and more like cloud compute with an agent native control layer. If Daytona keeps adding desktops, storage, and enterprise controls around the sandbox core, it can compete for a larger share of the budget that sits between model output and real work getting done.1 2