A powerful tool to turn your Windows PCs into a private, high-performance AI cluster for local LLM inference, image generation, and embeddings.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install windows-ai
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install windows-ai using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Windows AI Local Inference Cluster?
Windows AI leverages the ollama-herd project to transform standalone Windows machines into a unified local AI powerhouse. By utilizing Openclaw Skills, users can run large language models, generate images, and compute embeddings entirely on their own hardware without relying on expensive cloud APIs or subscriptions. This skill allows you to distribute AI workloads across multiple Windows devices—such as gaming rigs, work desktops, and laptops—creating a resilient and private AI infrastructure that maximizes your existing NVIDIA GPU resources via CUDA. It is designed for developers and enthusiasts who demand privacy, zero latency, and no per-token costs for their Windows AI workflows.
Windows AI Local Inference Cluster Use Cases
- Running LLMs like Llama 3 and DeepSeek privately on local Windows hardware.
- Building a local AI cluster by pooling GPU resources from multiple Windows PCs.
- Developing and testing AI-powered applications with an OpenAI-compatible endpoint.
- Generating high-quality images and text embeddings without per-token costs or rate limits.
How Windows AI Local Inference Cluster Works
- The user installs the Windows AI router (ollama-herd) on a primary machine to act as the central hub.
- Individual Windows machines are registered as nodes using the herd-node command to join the local cluster.
- The system implements 7-signal scoring and 15 health checks to monitor node performance in real-time.
- When an AI request is received, the vRAM-aware router identifies the best available GPU in the fleet to process the task.
- The response is served through a unified API endpoint compatible with standard OpenAI SDKs and popular AI tools.
Windows AI Local Inference Cluster Setup
To begin using Windows AI with Openclaw Skills, install the router and register your nodes via PowerShell:
# Install the core routing package
pip install ollama-herd
# Start the central router (default port 11435)
herd
# Register the local machine or additional network nodes
herd-node
# Configure Windows Firewall to allow traffic
netsh advfirewall firewall add rule name="Windows AI" dir=in action=allow protocol=tcp localport=11435
Windows AI Local Inference Cluster Data Schema & Taxonomy
The skill maintains its operational state and performance logs within the fleet manager directory. Below is the primary data structure used by the Openclaw Skills integration:
| Path | Purpose |
|---|---|
~/.fleet-manager/latency.db |
SQLite database tracking network and inference latency between nodes. |
~/.fleet-manager/logs/herd.jsonl |
Structured JSON logs for request tracking and error reporting. |
OLLAMA_KEEP_ALIVE |
Environment variable to control model residency in vRAM. |
OLLAMA_MAX_LOADED_MODELS |
Environment variable to manage concurrent model loading on a single node. |
Windows AI Local Inference Cluster Advanced Features
- vRAM-Aware Routing: Intelligently assigns models to GPUs with sufficient memory capacity to ensure optimal inference speed.
- Transparent Failover: Automatically retries failed requests on alternative Windows AI nodes if a specific machine goes offline.
- OpenAI Compatibility: Drop-in replacement for any tool supporting the OpenAI API specification, including Cursor and Continue.dev.
- Web Dashboard: Built-in monitoring available at
http://localhost:11435/dashboardfor real-time fleet visualization and health status. - Custom Tagging: Support for request tagging to track resource usage and performance metrics across different local projects or agents.
SKILL.md
Loading
Related Openclaw Skills
wsl2-local-ai
A high-performance LLM development stack for Windows users utilizing WSL2, NVIDIA GPU passthrough, and Ollama Herd routing.
photo-alchemy
A sophisticated AI pipeline that reimagines your personal photos as surrealist masterpieces using a multi-model workflow and deep macOS integration.
WeChat Message and File Sender
A comprehensive communication skill for Openclaw Skills that enables AI agents to query WeChat sessions and actively send text messages or various file types to contacts.
podcast-intel
Turns your Overcast listening history into a structured knowledge base to surface AI-driven insights from recent episodes.
Windows Ollama Fleet Manager
An advanced orchestration tool for routing Ollama inference requests across a network of Windows PCs to maximize hardware utilization.
RTX Local AI
RTX Local AI turns your gaming PC into a high-performance local AI server, allowing you to run massive LLMs with zero cloud costs using NVIDIA RTX GPUs.
Featured*
