Toolofficial publisher

Modal Sandboxes

Tool : A precise action the AI can call · requires modal

Serverless sandboxes with on-demand GPU.

The go-to choice when an agent needs to run heavy computation.

Docs

Overall score

6.1 /10

Security: one caveat, see the card

Utility

7.0

Reliability

6.0

Security

One caveat

Ease

Technical · Technical

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

Running code generated by a language model or untrusted user code on your own machine carries real risks. Modal Sandboxes addresses this by providing isolated containers, created on demand, in which any command can be executed without endangering the local environment.

In practice, you instantiate a Sandbox via the Python API (or the JavaScript and Go SDKs), pass it shell or Python commands, read the outputs in real time, then terminate it. Each Sandbox moves through well-defined lifecycle states: created, scheduled, started, ready, finished. You can mount volumes, open network tunnels, attach GPUs, and configure readiness probes to know exactly when the container is ready to accept traffic.

This tool is aimed at developers building AI agents, automated test pipelines, or user code execution environments. It also suits those who want to run arbitrary dependencies or setup scripts without polluting their base image.

On the other hand, it is not suited for use cases without Python skills or without a need for isolation: for simple tasks involving no untrusted code, the complexity of a serverless Sandbox is not warranted.

What it offers

·On-demand creation of isolated containers via Sandbox.create, with no infrastructure to manage.
·Execution of arbitrary commands (shell, Python, etc.) inside the Sandbox with real-time output reading.
·On-demand GPU support for compute-intensive tasks (paid).
·Observable lifecycle with distinct states: Created, Scheduled, Started, Ready, Finished.
·TCP or Exec readiness probes to automatically detect when the Sandbox is ready.
·Configurable timeout up to 24 hours, with an idle timeout to automatically stop an unused Sandbox.
·Multi-language compatibility: Python, JavaScript, and Go SDKs available.
·Filesystem snapshots to preserve a Sandbox's state and restore it in a new container.

Our tips

Use App.lookup with the create_if_missing=True parameter to automatically create an App if it does not yet exist, avoiding an error at startup.
Prefer a TCP probe (Probe.with_tcp) when your Sandbox starts a server, and an Exec probe (Probe.with_exec) for any other readiness condition such as the presence of a file or the completion of a setup script.
Configure an idle_timeout so the Sandbox stops automatically when there is no active command, stdin stream, or open TCP connection, to avoid unnecessary costs.
If your task exceeds 24 hours, use Filesystem Snapshots to save the Sandbox state and resume from a new container restored from that snapshot.
Call sb.terminate() explicitly at the end of your scripts to release resources immediately, rather than waiting for the timeout to expire.

Watch out

The maximum lifetime of a Sandbox is 24 hours; beyond that, Filesystem Snapshots are required to preserve state, which adds complexity.
A Sandbox created from outside a Modal container requires an App object to be explicitly passed to it, which can be surprising for users coming from the classic Function interface.

In practice

Price: Freemium
Commercial use: Commercial use: yes
Available on: sdk-python
Host: modal
FR interface: No

Security and data

Rating: Security: one caveat, see the card
Capabilities: readswritesexecutesnetwork
Your data: Data with the vendor
Maintenance: Active

Links and repository checked on 25 October 2026 · added on 1 January 2022

How to install Modal Sandboxes

1. Create an account on modal.com and retrieve your API token. 2. Install the Python SDK with: pip install modal 3. Authenticate with: modal token new (never share this token). 4. In your script, retrieve or create an App with modal.App.lookup("my-app", create_if_missing=True). 5. Create a Sandbox with modal.Sandbox.create(app=sb_app) and run your commands via sb.exec(...). 6. Always terminate the Sandbox with sb.terminate() to release resources immediately.

pip install modal

Official documentation ↗

In the same category

Whole category Code →