AIBnB
Download · Public beta

Get AIBnB.

One install: use big models over an OpenAI-compatible API — and, if you like, share a spare GPU with the network.

Choose your platform

Download · one package per platform
macOS
Apple Silicon · .pkg installer
No native installer yet — use the script install:
Run in your terminal:
curl -fsSL https://aibnb.tech/downloads/client-install.sh | bash
Windows
.exe native installer · no WSL
No native installer yet — use the script install:
Run in your terminal:
# Run inside WSL2 (Ubuntu):
curl -fsSL https://aibnb.tech/downloads/client-install.sh | bash
Linux
.deb / .rpm packages
No native installer yet — use the script install:
Run in your terminal:
curl -fsSL https://aibnb.tech/downloads/client-install.sh | bash
Prefer the script route, or want to read it first? client-install.sh.

One app. More possibilities.

  • OpenAI-compatible API — ready on install. Point your tools at it, or open the built-in web chat.
  • Plain text in, plain text out. A thin client; no GPU needed just to use it.
  • Your conversation data never touches the platform. It flows directly between nodes.
  • Spare GPU? Turn on Contribute compute in the app to carry a slice of the network — no separate install, nothing to wire up by hand.
  • Runs on macOS, Windows (native — no WSL) , and Linux.
Your device, part of a shared network.
Concept illustration
Manual install & self-check
  1. Clone the source (with submodules): git clone --recurse-submodules <repo> ~/.aibnb/src
  2. Install from the local path: uv tool install ~/.aibnb/src/client
  3. Generate config → aibnb-serve doctor (checks deps & reachability) → start (gives you aibnb-chat / aibnb-serve).
Server / advanced Standalone host install — headless nodes & servers

Most people don't need this. To share your GPU, just install AIBnB above and turn on Contribute compute — the app pulls down and sets up everything on demand, no separate host to manage. The standalone host package is only for headless servers and unattended machines with no desktop app: install once, run one segment of the model as a background service.

Needs: an Apple Silicon Mac (MLX) or an NVIDIA GPU + WSL2 (CUDA); more VRAM carries more layers, and disk must hold the assigned layer weights.

Standalone host · one package per system
macOS
Apple Silicon · .pkg installer
No native installer yet — use the script install:
Run in your terminal:
curl -fsSL https://aibnb.tech/downloads/host-install.sh | bash
Windows
.exe bootstrapper · auto-configures WSL2
No native installer yet — use the script install:
Run in your terminal:
# Run inside WSL2 (Ubuntu):
curl -fsSL https://aibnb.tech/downloads/host-install.sh | bash
Linux
.deb / .rpm packages · includes WSL2
No native installer yet — use the script install:
Run in your terminal:
curl -fsSL https://aibnb.tech/downloads/host-install.sh | bash
Prefer the script route? host-install.sh.
Manual install & self-check
  1. Clone the source (with submodules): git clone --recurse-submodules <repo> ~/.aibnb/src
  2. Install from the local path (uv resolves ../proto): uv tool install ~/.aibnb/src/host
  3. Pre-flight & run: aibnb-host doctor checks platform / GPU / VRAM / disk / deps / reachability, then the service unit (launchd / systemd, Restart=always) starts it.
Beta builds are unsigned — Gatekeeper / SmartScreen will warn; follow the note on each card, and verify the sha256 before installing. Signing, notarization, and auto-update are planned for a later stage.