Kern Sandbox, run LLM-generated code locally, no cloud, no daemon
Runs LLM-generated code in disposable local containers — no cloud, daemon, or account.
Kern Sandbox executes AI-generated code in a fresh OCI container with no network, memory and PID caps, and an external deadline, then destroys the container — so a hallucinated `rm -rf ~` resolves to the container's own read-only root, not your machine. It's a static binary plus a Python API (`kern_sandbox`), with no daemon, socket, VM, cloud or account required. You can keep state across calls with a persistent Sandbox() or a warm kernel() interpreter, and standard-library imports are precompiled per image for about 3x cheaper cold starts.
What holds up
- +No network, memory or PID limits by default — you open only the hosts you name with egress_allow.
- +No daemon, socket, VM, cloud or account — just a static binary plus a Python package.
- +Container is thrown away after each call, so a hallucinated rm -rf ~ hits a disposable read-only root.
- +Standard-library imports are precompiled per image, making a fresh container's imports about 3x cheaper.
Mind the limits
- −Setup needs the native binary (curl script) plus the Python package — two installs, not one.
- −Each run_code() call spins up a new container by default — avoiding that overhead needs the kernel() mode.
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/kern-sandbox-run-llm-generated-code-locally-no-cloud-no-daem/"><img src="https://stillworks.dev/badge/kern-sandbox-run-llm-generated-code-locally-no-cloud-no-daem.svg" alt="Picked by StillWorks" width="230" height="54"></a>