Kern Sandbox runs LLM-generated code and agent tool calls rootless and daemonless, in kernel-enforced isolation, from a CLI or Python/Node SDKs.
Yes, Kern Sandbox is open source under the Apache-2.0 license.
Kern Sandbox is free to use.
Yes, Kern Sandbox can be self-hosted (the source is available under the Apache-2.0 license).
Kern Sandbox is a rootless, daemonless container runtime that runs LLM-generated code and agent tool calls in kernel-enforced isolation, callable from a CLI or Python/Node SDKs with MCP support. Open source: yes (Apache-2.0); self-hostable; pricing free. It ships as a single static binary and has 427 stars.
Kern Sandbox gives an agent somewhere safe and fast to run code. It executes LLM-generated code and tool calls inside a container isolated by the kernel, but without a root requirement and without a background daemon — the whole thing is a single static binary you drop in and call. You reach it from a CLI, embed it through the Python or Node SDK, or expose it to an agent over MCP. The design target is throughput at agent scale: a prewarm pool keeps ready sandboxes on hand so a tool call does not pay full container-startup cost each time. Open source: yes (Apache-2.0); self-hostable; pricing free.
The moment a GTM agent starts running code — parsing a scraped page, transforming an enrichment payload, executing a generated SQL or Python snippet — you need that code contained, and you need it cheap enough to do thousands of times a day. Kern Sandbox targets exactly that: rootless, daemonless isolation you can self-host, addressable over MCP so an agent can call it as a tool, with a prewarm pool aimed at keeping per-call cost low. Open source: yes (Apache-2.0); self-hostable; pricing free. The honest read: the isolation is a container boundary rather than a VM, and the eye-catching cost and speed multiples are the vendor's own benchmarks — pilot it against your real tool-call mix and confirm both the containment guarantees and the numbers before it fronts untrusted execution at scale.