Fork My Stack

My entire AI assistant setup. Yours in an afternoon.

Quick Start • Architecture • Guides • Cost • Blog

---

🚀 What This Is

fork-my-stack is my production-style starter stack for running a personal AI assistant with real workflows, real memory, and real automation.

You get:

• Channel-ready assistant patterns (WhatsApp, Telegram, Slack, Discord)
• Multi-model routing (frontier + coding + local Ollama)
• Cron automations for recurring tasks
• MCP integrations for external systems/tools
• Structured memory system (daily + long-term)
• Sub-agent execution model for isolated project work

---

⚡ Quick Start

# 1) Clone
cd ~/Development
git clone https://github.com/YOUR_USERNAME/fork-my-stack.git
cd fork-my-stack

# 2) Copy workspace scaffold
mkdir -p ~/.openclaw/workspace
cp -R workspace/* ~/.openclaw/workspace/

# 3) Create gateway config from template
mkdir -p ~/.openclaw/config
cp config/gateway-template.yaml ~/.openclaw/config/gateway.yaml

# 4) Edit config (models, channels, MCP servers, keys)
$EDITOR ~/.openclaw/config/gateway.yaml

# 5) Start OpenClaw
openclaw gateway start
openclaw gateway status

# 6) Validate
openclaw help

If using local inference, set up Ollama via guides/ollama-setup.md.

---

🧠 Architecture

You (any device) → Channels (WhatsApp/Telegram/Slack/Discord) → OpenClaw Gateway → Agent Loop
                                                                        ↓
                                                    ┌─────────────────────────────────┐
                                                    │  Frontier Model (Opus/GPT-4)    │
                                                    │  Coding Model (Codex)           │
                                                    │  Local Model (Ollama, $0)       │
                                                    └─────────────────────────────────┘
                                                                        ↓
                                                    MCP Tools │ Cron Jobs │ Memory │ Sub-agents

---

📦 What's Inside

Path	Type	Purpose
README.md	File	Main setup + architecture + docs index
LICENSE	File	MIT license
.gitignore	File	Git ignore rules
assets/	Directory	README images (banner + architecture)
config/	Directory	OpenClaw gateway config templates
config/gateway-template.yaml	File	Starter gateway config
workspace/	Directory	Assistant identity, memory, and operating docs
workspace/AGENTS.md	File	Global operating rules and routing policy
workspace/SOUL.md	File	Assistant personality and behavior profile
workspace/USER.md	File	User-specific preferences/context
workspace/TOOLS.md	File	Environment/tool notes
workspace/MEMORY.md	File	Long-term curated memory
workspace/HEARTBEAT.md	File	Periodic check instructions
workspace/memory/	Directory	Daily memory/log files
guides/	Directory	Practical setup and strategy docs
guides/ollama-setup.md	File	Local model setup
guides/multi-model-routing.md	File	Model routing strategy
guides/mcp-servers.md	File	MCP server catalog
guides/blog-automation.md	File	Content automation workflows
crons/	Directory	Scheduled automation job examples
crons/channel-monitor.json	File	Channel monitoring automation
crons/weekly-build-log.json	File	Weekly build-log automation
crons/daily-content-pitches.json	File	Daily idea/content automation
crons/memory-maintenance.json	File	Memory upkeep automation

---

💸 Cost Breakdown

Component	Free Tier / Self-Hosted	Paid Option
OpenClaw runtime	✅ Run locally	Optional hosted infra
Ollama local model	✅ $0/token on your hardware	GPU cloud if you outsource
Frontier model (Opus / GPT-4)	❌	API usage based
Coding model (Codex-family)	❌	API usage based
Messaging channels	⚠️ Some free/dev tiers	Provider/API pricing
MCP servers	✅ Self-host OSS servers	Managed services / external APIs

Rule of thumb: use frontier models for quality-critical work, coding models for implementation, and Ollama for low-cost recurring jobs.

---

🔌 MCP Servers (11)

Server	Repo
mcp-proxmox-admin	https://github.com/EricGrill/mcp-proxmox-admin
mcp-kali-orchestration	https://github.com/EricGrill/mcp-kali-orchestration
mcp-bitcoin-cli	https://github.com/EricGrill/mcp-bitcoin-cli
mcp-multi-agent-ssh	https://github.com/EricGrill/mcp-multi-agent-ssh
mcp-multi-agent-server-delegation	https://github.com/EricGrill/mcp-multi-agent-server-delegation
mcp-predictive-market	https://github.com/EricGrill/mcp-predictive-market
mcp-civic-data	https://github.com/EricGrill/mcp-civic-data
mcp-memvid-state-service	https://github.com/EricGrill/mcp-memvid-state-service
mcp-market-data	https://github.com/EricGrill/mcp-market-data
mcp-ipfs	https://github.com/EricGrill/mcp-ipfs
quickbooks-online-mcp-server	https://github.com/EricGrill/quickbooks-online-mcp-server

---

📚 Guides

• Ollama Setup
• Multi-Model Routing
• MCP Servers
• Blog Automation

---

✍️ Blog Posts

Related reading from ericgrill.com:

• PicoClaw vs OpenClaw: A Code-Level Deep Dive
• What You Lose (and Gain) Running PicoClaw Instead of OpenClaw
• Running an AI Agent on a $10 Board vs a Mac Mini
• Run Your AI Assistant on a Local GPU for (Basically) Free
• Claude Code Agents
• Claude Code Hooks
• Claude Code Skills
• Proxmox Cluster Rescue with Claude Code
• QuickBooks MCP: AI Bookkeeper
• Memvid: Portable AI Memory

---

👋 About

Built by Eric Grill - Builder | Breaker | Fixer.

• Website: https://ericgrill.com
• GitHub: https://github.com/EricGrill
• X/Twitter: https://x.com/EricGrill

If this helped you, fork it, customize it, and ship your own assistant stack.