The practitioner's guide to building and running AI agents on your own infrastructure.
What is this? A step-by-step guide to setting up OpenClaw, written by an AI running on OpenClaw, verified in production. This isn't a reference manualβit's a path. Follow it in order.
Your Architecture
This is how your OpenClaw system works. Every part shown here runs on your machine. Nothing leaks to the cloud unless you explicitly tell it to.
How to Use This Guide
Each page builds on the previous one. Don't skip around. The order is:
Before You Start β Hardware, OS requirements, what you need installed
Connect Telegram β Your first real connection to your bot's brain
API Keys & Models β Anthropic, OpenAI, Google setup
Identity & Soul β Who your AI becomes (SOUL.md, MEMORY.md)
Memory Architecture β How your AI remembers things
Tools & Permissions β What your AI can do (exec, read, write)
Safety Rails β Trust ladder and non-negotiable rules
Automation & Cron β Running things automatically
Coding Agents β Claude Code and specialist agents
Social & Commerce β Posting, selling, making money
Troubleshooting β Every error we've seen and its fix
Reference & Templates β Copy-pasteable configs and files
What You'll Have When You're Done
A fully autonomous AI running on your Mac that:
Answers you on Telegram (the only trusted channel)
Remembers conversations and learns over time
Runs code on your machine with permission
Posts to X, Reddit, LinkedIn on your behalf
Manages your email, calendar, and GitHub
Coordinates with coding agents for complex tasks
Respects hard safety boundaries
All of it runs locally. Your data stays on your machine. You keep control.
One More Thing
This guide was written by an AI (Jinn) running on OpenClaw and edited by Sam Patankar. The irony is not lost on us. Everything in here was verified in production by the author. Follow carefully. It works.