Yokai — Your AI fleet, under one command

GPU fleet control for self-hosted AI

Your AI fleet,
under one command.

Discover devices. Deploy best-known configs. Watch every token and watt.

Install Yokai Why Yokai exists

Open sourceSingle release archiveKeys stay local

YouOpenTUI

Yokai daemonFleet controller · healthy

GPU-BOX-01vLLM · running

GPU-BOX-02llama.cpp · running

HOMELAB-3090ComfyUI · idle

Fleet telemetryLive

GPUs online

3 / 3

Total VRAM

304 GB

Services

8 running

Power draw

1.32 kW

Avg. GPU util.

67%

01

QuickstartInstall in seconds.

$ curl -fsSL https://raw.githubusercontent.com/spencerbull/Yokai/main/install.sh | sh Copy

Detects macOS or Linux, installs both Yokai binaries, and restarts an existing daemon safely.

02

DocumentationEverything you need to ship and operate.

Get startedInstall Yokai and bring your first GPU online.

Operate your fleetDeploy, inspect, test, restart, and upgrade.

Integrate AI toolsWire local models into your everyday workflow.

Understand the systemFollow traffic from the TUI to every agent.

Built for operators

From scattered hardware
to one observable system.

1. DiscoverImport SSH hosts or find peers over Tailscale.
2. DeployApply GPU-aware recipes or tune every runtime flag.
3. ObserveInspect health, utilization, logs, and LLM throughput.
4. IntegrateRegister every endpoint with your coding tools.