A high-performance local AI inference engine that aggregates the compute power of M-series Macs into a unified, scalable cluster.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install apple-silicon-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 apple-silicon-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 Apple Silicon AI Fleet Manager?
Apple Silicon AI is a specialized toolset designed to harness the full potential of Apple unified memory architecture for local AI workloads. By utilizing Openclaw Skills, users can transform individual Mac Studio, Mac Mini, and MacBook Pro devices into a cohesive inference fleet. This setup bypasses the need for costly cloud-based GPU rentals or privacy-risking APIs by keeping data and processing entirely on-premise. The skill leverages the high-bandwidth memory of M1, M2, M3, and M4 chips to run massive models that typically require multiple enterprise-grade GPUs, making it a professional-grade solution for local LLM execution, image generation, and speech processing.
Apple Silicon AI Fleet Manager Use Cases
- Scaling local LLM inference across multiple Apple Silicon devices to support high-concurrency developer workflows.
- Generating high-resolution images natively using Flux models without relying on external cloud subscriptions.
- Transcribing audio locally with Qwen-ASR to ensure maximum data privacy for sensitive recordings.
- Building distributed RAG pipelines by batching document embeddings across several networked Mac devices.
How Apple Silicon AI Fleet Manager Works
- The user installs the core inference package on all target Apple Silicon machines within a local network.
- A single Mac is initialized as the cluster router, which serves as the primary endpoint for all Openclaw Skills requests.
- Additional Macs are started as node agents, which use auto-discovery to register their hardware capabilities with the central router.
- The system continuously monitors seven hardware signals—including thermal pressure, memory availability, and queue depth—to determine the most efficient node for every request.
- When an API call is made, the router transparently handles the model loading and task execution on the optimal device, returning the result to the user through a unified interface.
Apple Silicon AI Fleet Manager Setup
Install and configure your local cluster using the following commands:
# Install the fleet manager on every Mac
pip install ollama-herd
# Start the router on your primary Mac (defaults to port 11435)
herd
# Join other Macs to the fleet as worker nodes
herd-node
Once started, the cluster is accessible via OpenAI-compatible endpoints at http://localhost:11435/v1.
Apple Silicon AI Fleet Manager Data Schema & Taxonomy
Apple Silicon AI manages system state and performance metrics locally using a structured file hierarchy. No model data or user inputs are sent to external servers.
| Data Component | Path | Description |
|---|---|---|
| Routing Metrics | ~/.fleet-manager/latency.db |
SQLite database tracking response times and node performance. |
| System Logs | ~/.fleet-manager/logs/herd.jsonl |
Structured JSON lines documenting request lifecycle and errors. |
| Model Storage | /usr/share/ollama/.ollama |
Standard storage for downloaded LLM weights and configurations. |
| Metadata | Internal Registry | Real-time map of node hardware (RAM, chip gen) and loaded models. |
Apple Silicon AI Fleet Manager Advanced Features
- Intelligent load balancing that routes requests based on real-time thermal state and memory pressure of individual Macs.
- Native support for MLX-optimized models including Flux for image generation and Qwen-ASR for speech-to-text.
- Real-time fleet dashboard accessible via browser for monitoring model distribution and hardware health.
- Hardware-aware model recommender that suggests the best parameter sizes (7B, 32B, 70B, etc.) based on available unified memory.
- Fully local, API-key-free environment ensuring complete data sovereignty for enterprise or private development.
SKILL.md
Loading
Related Openclaw Skills
Automation Suggestion
An intelligent system that analyzes manual task patterns to propose high-value automations without manual intervention or nagging.
Amazon PPC
A comprehensive framework for architecting and optimizing Amazon Sponsored Ads based on margin-driven logic and tiered campaign structures.
Amazon SEO
A comprehensive optimization framework for Amazon product listings designed to increase organic rank and conversion rates through strategic keyword placement.
TradingView Technical Indicators Analysis Assistant
A sophisticated toolset that converts 100+ popular TradingView Pine Script indicators into Python-accessible technical analysis functions.
Amarin Memory
A lightweight, SQLite-powered memory system for AI agents that handles long-term storage, semantic retrieval, and natural memory decay.
Loyalty Designer
A strategic framework for designing margin-positive loyalty programs that increase customer lifetime value through points, tiers, and referral mechanics.
Featured*
