Together Code Sandbox

What it does

Together Code Sandbox takes a Docker image or a Dockerfile build context, turns it into a snapshot, and starts isolated virtual machines from that snapshot. Through a Python SDK, a TypeScript SDK or a CLI, you run commands, read and write files, open terminals and get preview URLs for services running inside the VM. The output is command results, file contents and live preview links, plus saved snapshots you can resume later.

Teams use it to execute untrusted or model-generated code, run tool calls from agents, and provide rollout environments for reinforcement learning. Its distinguishing features are memory-and-disk snapshotting, with Together citing sub-second resume and cloning, and Dev Container support for servers, databases and custom Docker setups. VMs scale from 2 to 64 vCPUs and 1 to 128 GB RAM, and files persist on a git-versioned filesystem.

It suits developers building AI products, coding agents or evaluation and training pipelines who want managed VMs instead of building their own. You need a Together API key and an organization with the SDK and CLI enabled, which the docs say requires contacting sales. It is programmatic only, with no standalone app, and billing is metered per vCPU and per GiB of RAM per hour.

What it can do

• Build snapshots from public Docker images or Dockerfile build contexts using a remote image builder
• Create sandboxes from snapshots and run commands or interactive terminals
• Hibernate and resume sandboxes with disk and memory state preserved
• Resize VMs from 2 to 64 vCPUs and 1 to 128 GB RAM
• Read, write and watch files on a persistent git-versioned filesystem
• Expose preview URLs for services running inside a sandbox

Example workflows

Run model-generated code safely

Input: a public Docker image such as python:3.12-slim. Steps: create a snapshot with an alias, create a sandbox from that alias, run the generated code with the command API, then terminate the sandbox. Output: the command's output and any files written inside the VM.

Preview an AI-built web app

Input: a project template or Dockerfile context. Steps: create a sandbox, run the install and build commands, request the preview URL for the app's port, then hibernate the sandbox. Output: a live preview link and a saved snapshot to resume or fork later.

Before you choose

• Access to the SDK and CLI is gated per organization and requires contacting sales
• Sandbox and Code Interpreter are listed as separate items on the pricing page; the Code Interpreter's price was not captured, so check which one applies
• Marketing speed figures (cold start, snapshot resume) are vendor claims and may differ from your workloads

Listed under

• Coding
• Sandboxes and infrastructure
• AI apps
• AI developers
• Code execution
• Code interpreter
• Developer tools
• Dev environments

Get started

Org must have SDK/CLI enabled via sales; auth uses TOGETHER_API_KEY

1. Ask Together sales to enable the sandbox SDK and CLI for your organization
2. Install together-sandbox for Python or TypeScript, or run the CLI installer, and set TOGETHER_API_KEY
3. Create a snapshot from a Docker image or build context, then create a sandbox from its alias

Code Sandbox is metered per hour by vCPU and by GiB of RAM. The pricing page does not state billing granularity or a free allowance.

Source: together.ai/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Sandbox | Together AI ↗www.together.ai
• Pricing | Together AI ↗www.together.ai
• Code sandbox - Together AI docs ↗docs.together.ai

Company website

together.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.