Every machine you own, one endpoint. XBANQ Cluster sits on top of the Ollama installs you already have and turns them into a single OpenAI-compatible API — routed to whichever machine can actually run the model, with live monitoring of every machine in the fleet.
What you get
- One OpenAI-compatible API endpoint for your whole fleet — drop-in for Cursor, LangChain, Continue and any OpenAI client
- VRAM-aware routing: each request lands on a machine that can actually run the model
- Circuit-breaker failover and automatic retry when a machine drops out
- Live traffic and 3D fleet monitoring: CPU, RAM, VRAM, GPUs and models per machine
- Intent-based dispatch — pick an objective such as FASTEST, CHEAPEST or PRIVATE and the router follows it
- Trust tiers, jurisdiction and policy controls (Compute Fabric) — keep workloads on the machines you trust
- Model placement advisor — see which machine should serve which model before you pull it
- Chat playground with attachments and token-by-token streaming
- mTLS between coordinator and machines, hashed API keys, rate limiting, audit log
- Interface in 13 languages
- 14-day free trial with 5 machines, then keep the free tier (2 machines) forever — no card required to start
Honest scope
Each request runs on exactly one machine. XBANQ Cluster does not split a model across machines — it makes the machines you already own behave like one reliable, monitored service. No cloud, no telemetry of your prompts: inference stays on your hardware.
Every feature runs on every tier. The licence sets how many machines your cluster may count — nothing else. SSO and audit export for Business customers are in build and marked as coming.
How it works
- Install Ollama on each machine (if it is not there already)
- Start the coordinator on your main machine — the desktop app does it with one click
- Join every other machine with one command
- Point your tools at http://your-cluster:4317/v1 — done
System requirements
- macOS 13 or newer for the desktop app (desktop builds ship for macOS today)
- Linux for the headless coordinator/server bundle
- Node.js 20+ on the coordinator machine
- Ollama on every machine that joins the cluster
- Redis recommended on the coordinator (persistence, caching, distributed rate limiting)