One command
Install with a single line, sign in with Google, and your coding agent is linked automatically. See Install.
Waired documentation
Your own AI, on your own hardware — reachable from anywhere over a private, encrypted link.
Waired is a private, inference-only overlay network. It connects your laptops, GPU servers, VMs, and home boxes so that local LLMs can be shared across your own devices without exposing them to the public internet. Type in Claude Code on any laptop and the answer comes from your own machine — no per-message bill, no subscription, and your prompts never leave your devices.
Free to run
The model runs on hardware you already own — a recent CPU handles a 7B model, and a GPU makes it fast.
Private by design
Peers talk directly over an end-to-end encrypted WireGuard link. The coordination service never sees your prompts or replies — even the relay can’t read them. See Privacy.
Your tools, unchanged
Works with Claude Code and OpenCode out of the box, plus any OpenAI- or Anthropic-compatible client.
Start here
Section titled “Start here”InstallOne-line install for Linux, macOS, and Windows.
First runEnroll a device, enable inference, and pull your first model.
Coding agentsRoute Claude Code and OpenCode through your own model.
Model catalogBundled models and the RAM / VRAM each one needs.
CLI referenceEvery `waired` command, at a glance.
TroubleshootingThe three commands to run when something looks wrong.