> ## Documentation Index
> Fetch the complete documentation index at: https://docs.katakate.org/llms.txt
> Use this file to discover all available pages before exploring further.

# K7d

> Fork a running Kubernetes cluster in ~100 ms. Run 50 copies on one 64 GB box.

<Frame>
  <img src="https://mintcdn.com/katakate/_XWVDYhC_wp6k1ZI/k7d/k7d-cover.png?fit=max&auto=format&n=_XWVDYhC_wp6k1ZI&q=85&s=c72a0ec4ef9bb72856e9ce63bb686668" alt="k7d: fork isolated Kubernetes clusters in ~100 ms" width="1280" height="720" data-path="k7d/k7d-cover.png" />
</Frame>

k7d is a Rust VMM that forks running VMs — including a live multi-VM Kubernetes cluster — in \~100 ms, with memory, disk, processes, and network identity intact. Boot once, then explore dozens of byte-identical worlds for RL and agent evals. For sandboxes at scale with a CLI / API / Python SDK, see [K7](/k7/index).

<CardGroup cols={2}>
  <Card title="CoW fork" icon="bolt" href="/k7d/concepts/cow-fork">
    How a 5 ms VM fork works: shared memory until write, dirty pages, byte-identical copies.
  </Card>

  <Card title="GRPO / tree search" icon="sitemap" href="/k7d/guides/grpo">
    Wire k7d into a trainer: fork\_batch, protect, prune, auto\_evict under RAM budgets.
  </Card>

  <Card title="Cluster mode" icon="network-wired" href="/k7d/guides/cluster-mode">
    Create or adopt an N-node k3s cluster and fork it without agent restarts.
  </Card>

  <Card title="Daemon API" icon="code" href="/k7d/api/protocol">
    JSON-lines protocol on /run/k7d/k7d.sock — every tree and VM verb.
  </Card>

  <Card title="Benchmarks" icon="gauge-high" href="/k7d/deep-dives/benchmarks">
    Hardware, exact commands, and every enforced latency budget.
  </Card>

  <Card title="Formal verification" icon="shield-halved" href="/k7d/deep-dives/formal-verification">
    Kani on unsafe paths; Aeneas→Lean on the tree budget model.
  </Card>
</CardGroup>