Isolated cloud sandboxes for running AI-generated code and agent tool calls

Daytona addresses a problem that arrives with every autonomous agent: the code the model produces has to run somewhere, and running it on application infrastructure hands an unreviewed program the same privileges as the application. Daytona's answer is a sandbox primitive exposed through an API and SDKs for Python and TypeScript. A sandbox is created on demand, isolated from other tenants, and torn down when the task ends. Inside it an agent can execute processes and stream their output, read and write files under permission controls, run Git operations with credentials that stay outside the sandbox, and open an SSH or browser-based VS Code session when an engineer needs to inspect what happened.

The platform supports Python, TypeScript, Ruby, Go, and Java, along with Dockerfiles, Docker Compose, and Docker-in-Docker for workloads that carry their own images. Snapshots capture sandbox state so long-running or interrupted workflows resume rather than restart, and Computer Use sandboxes extend the same model to full Linux, macOS, and Windows desktops for agents that drive a GUI. Customers can run on Daytona's compute or on their own. The company says provisioning completes in under 90 milliseconds. Daytona was founded in 2023 by Ivan Burazin, Vedran Jukic, and Goran Draganic, is based in New York, and has raised roughly $31M including a $24M Series A led by FirstMark Capital.