Skip to main content
k7d: fork isolated Kubernetes clusters in ~100 ms
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.

CoW fork

How a 5 ms VM fork works: shared memory until write, dirty pages, byte-identical copies.

GRPO / tree search

Wire k7d into a trainer: fork_batch, protect, prune, auto_evict under RAM budgets.

Cluster mode

Create or adopt an N-node k3s cluster and fork it without agent restarts.

Daemon API

JSON-lines protocol on /run/k7d/k7d.sock — every tree and VM verb.

Benchmarks

Hardware, exact commands, and every enforced latency budget.

Formal verification

Kani on unsafe paths; Aeneas→Lean on the tree budget model.