rbee Documentation

rbee is a self-hosted AI orchestration system that turns every GPU you own—across desktops, laptops, servers, and remote machines—into a single, unified AI cloud. This documentation is for users building and operating rbee systems, from single-machine setups to multi-device fleets.

Who this documentation is for

Homelab & power users

Run AI models locally across your gaming PC, laptop, and home server. Use all your GPUs together in one interface without juggling ports or manual configuration. Currently supports text inference (LLMs), with multi-modal support (images, audio, video) planned for M3 (Q1 2026).

Start here: Getting started: single machine or Getting started: homelab

GPU providers & businesses

Turn your GPU fleet into an API product. Expose heterogeneous hardware through one OpenAI-compatible endpoint with production-grade routing, telemetry, and scheduling.

Start here: Getting started: GPU providers or See Premium Modules for advanced routing, telemetry, and GDPR-focused auditing (€129-€499 one-time, planned for M2 launch Q2 2026).

Academic & research teams

Deploy on-premises AI infrastructure with data residency guarantees. Use GDPR-focused auditing modules for compliance-sensitive workloads.

Start here: Getting started: academic & research or GDPR & compliance

Quickstart paths

• Single machine setup - Run rbee on one computer with one or more GPUs
• Multi-machine homelab - Connect multiple machines into one unified system
• GPU provider platform - Build an API product from your GPU fleet
• Licensing & premium modules - Understand what’s free vs premium

Core concepts

rbee uses a distributed architecture with four main components:

• Keeper - The GUI application where you control everything. Runs on your workstation.
• Queen - The orchestrator that routes requests and manages the system. One per colony.
• Hive - A host machine (physical or virtual) that runs workers. You can have many hives.
• Worker - An inference process running a specific model. Currently supports LLM inference. Each worker uses specific GPU resources.

Together, these form a colony - your unified AI system.

Learn more: Architecture overview

Licensing & premium modules

rbee has a mixed licensing model:

• Open core - The base system (keeper, queen, hive, basic workers) is open source under GPL-3.0 and MIT licenses. Free to run forever, no feature limits for personal use.
• Premium binaries - Production-grade modules (Premium Queen, Premium Worker, GDPR Auditing) are proprietary and sold as one-time lifetime licenses. No subscriptions.

Premium modules add advanced routing, deep telemetry, quota management, and compliance features for businesses running rbee as a product.

All premium features are clearly marked in the documentation.

Learn more: Licensing details or Premium modules overview

Where to go next

1. Install rbee - Download and set up the base system
2. Run your first colony - Get a single-machine setup working
3. Scale to multiple machines - Connect your devices into one system
4. Use the OpenAI-compatible API - Integrate rbee into your applications
5. Understand GDPR & auditing - Learn about compliance features