Skip to content
where do irun this.

A guide to choosing where to run

Code & AI agent sandboxes

Compare code sandboxes by isolation boundary, environment lifetime and assessed capabilities before choosing where an agent runs code.

What to check before choosing

Start with the code you need to contain. Running untrusted commands calls for checking the documented isolation boundary, network controls and access to secrets. A sandbox label alone is not a security guarantee; the atlas records the boundary a product describes rather than ranking its security.

Then check the execution lifecycle: supported languages, setup time, session limits, filesystem persistence and how outputs leave the environment. An interactive agent loop and a short batch job can need very different lifetimes and interfaces.

This list includes every product classified as sandbox, not just products with verified agent-harness integrations. It excludes products classified as persistent environments and pure substrate libraries. Check each profile for evidence of the specific harness and capabilities you need; use the full atlas to explore other product types.

How we assess products · Explore this type in the atlas

Products in this category

where do i run this.

An independent spec sheet for places to run code, and the code your agents write. Facts come from each product's public pages; every capability is a count, never a grade.

Verify

Pages change; so do prices. Verify on the vendor's site before you commit; every product page shows when it was read.

Where Do I Run This © 2026counts, not grades · public pages only