Novita Agent Sandbox provides isolated, stateful execution environments for AI agents.
You can use a Sandbox to run code, install dependencies, access files, use browsers, and preserve execution state across sessions without managing infrastructure.
Core Concepts
Agent Sandbox is built around three core concepts: Sandbox, Template, and Snapshot.
Sandbox
A Sandbox is an isolated execution environment where your agent performs tasks.
Inside a Sandbox, your agent can run commands, read and write files, install dependencies, and use browser-based workflows. Each Sandbox has its own filesystem and runtime environment.
Template
A Template defines the environment a Sandbox starts with.
Templates help you create consistent Sandboxes with predefined dependencies, configuration, and setup steps. Use Templates when you want multiple Sandboxes to start from the same prepared environment.
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.
When to Use Agent Sandbox
Agent Sandbox is designed for workloads where agents need an isolated environment to execute tasks, keep state, or reuse prepared environments.
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
Pricing
Agent Sandbox provides per-second billing for CPU and RAM. Storage — encompassing templates, paused sandboxes, and snapshots — is now subject to daily charges. See the pricing page for more details.
Getting Started
You can start and manage your Agent Sandboxes through SDK or CLI. For details, please refer to the quick start page. Last modified on June 12, 2026
