Skip to main content
Novita Sandbox offers isolated, stateful execution environments designed for AI agents. It enables running code, installing dependencies, accessing files, using browsers, and preserving state across sessions without infrastructure management.
Novita Sandbox is a managed execution environment for AI agents that need to run code, use browsers, work with files, and preserve state across sessions. It gives each agent an isolated runtime without requiring you to manage infrastructure.
The platform is built on three foundational concepts:

​
Sandbox

An isolated execution environment where agents perform tasks. Each has its own filesystem and runtime, allowing agents to run commands, manage files, install dependencies, and handle browser-based workflows.

​
Lifecycle

Sandbox lifecycle: transitions between running, paused, and killed
A sandbox moves through three states: Running, Paused, and Killed.
  • Running — the sandbox is active and can execute commands and serve connections.
  • Paused — the sandbox is suspended; its filesystem and memory state are preserved and billing for compute stops. It can be resumed back to Running (typically in about a second).
  • Killed — the sandbox is terminated and its resources are released. A killed sandbox cannot be resumed.
Two settings drive the transitions out of Running:
  • Timeout — the maximum lifetime of the sandbox. It counts down from creation regardless of activity; when it expires the sandbox is stopped by default, or paused if you configure the lifecycle accordingly. See Timeout Handle.
  • Idle timeout — fires only when no client has been connected for the configured duration, automatically killing or pausing the sandbox. See Idle Timeout Handle.
On timeout or idle timeout you can choose pause (so the sandbox can be resumed later, optionally with auto-resume) instead of kill. When paused, resuming restores the sandbox to Running with its state intact; when killed, the sandbox and its data are gone.

​
Template

Defines the starting environment for a Sandbox. Templates ensure consistency across multiple Sandboxes with predefined dependencies, configuration, and setup steps.

​
Snapshot

A Snapshot saves the current state of a Sandbox as a reusable environment. Snapshots are useful after you have configured a Sandbox, installed dependencies, or generated files that you want to reuse. New Sandboxes can be created from a Snapshot without repeating the same setup steps.
Intended for workloads requiring isolated environments to execute tasks, maintain state, or reuse prepared setups. Common use cases include:
  • Coding agents that run code, install packages, and execute tests
  • Browser agents that automate web workflows
  • Data analysis agents that process files and generate reports
  • Research or RL workloads that need many isolated environments
Billing is per-second for CPU and RAM. Storage—covering templates, paused sandboxes, and snapshots—incurs daily charges. Additional details are available on the pricing page.
Sandboxes can be started and managed via SDK or CLI. The quick start page provides further guidance.
Last modified on August 11, 2026