daytonaio/daytona

Daytona is an infrastructure platform designed to safely run code generated by AI agents. The core problem it addresses is that when an AI system writes and executes code, it needs an isolated, secure environment so that the code cannot accidentally (or deliberately) harm the host machine or access sensitive data. Daytona provides these isolated environments, called sandboxes, which are essentially lightweight virtual computers with their own filesystem, network, CPU allocation, and memory, fully isolated from each other and from the host. Sandboxes spin up in under 90 milliseconds, which makes them fast enough for interactive AI agent workflows where code needs to be executed in near-real time. Developers and AI agents interact with sandboxes through SDKs available in Python, TypeScript, and JavaScript, or through a REST API and a command-line interface. Operations include running code, reading and writing files, managing processes, and taking snapshots (saved states that let an agent resume exactly where it left off in a later session). The platform is built on OCI and Docker compatibility, meaning it uses the same container standards as the broader DevOps ecosystem. Additional features include VNC and SSH access for human inspection, Git operations, webhooks, network controls, and a web dashboard. You would use Daytona if you are building an AI coding agent that needs to execute arbitrary code safely, running automated code generation pipelines, or providing a sandboxed code execution environment as part of a product. It is also useful for running isolated development environments. The tech stack is TypeScript and NestJS for the primary API, Go for the CLI, with sandboxes using containerization. The platform is open-source and also available as a hosted service.