One interface for access to the
Bumbary AI cluster.
The Bumbary AI Cluster Platform is a distributed, multi-agent AI system designed to run locally, on-premise with full control, observability, and collaboration capabilities. Users are able to chat with agents, compose workflows, monitor cluster resources (e.g. CPUs, GPUs, etc).
Supabase VPS cluster server… ETA 38s.Everything you need to use AI in a private compute cluster.
Built for a person who cares about data privacy, token cost, and performance.
Streaming Chat
Talk to your agents with token-by-token responses & voice input.
Agent Marketplace
Discover, enable & version plugins across your cluster.
Live Metrics
GPU, CPU, temperature & fans — every node, every second.
Visual Workflows
Drag-and-drop nodes to compose multi-agent pipelines.
Smart Scheduler
Cron-style jobs with retries, alerts and dependency graphs.
Audit Trail
Every action, every user — searchable and exportable.
Docs
Architecture & reference
Diagrams describing how the Bumbary local AI cluster fits together — from hardware tiers to LLM model placement on each tiered node.
Local AI Cluster Architecture
End-to-end view of control plane, orchestrator and GPU node pool.
Four-Tier Hardware
DGX Spark, Mac Studio, RTX agents and ingestion nodes.
Model Placement
Authoritative map of LLMs across each tier.
System Architecture
Final form: users → Mac → orchestrator → specialists.
Logical + Physical
How logical roles map to physical nodes.
Platform Features
Chat, agents, workflows, observability and security.