What Is This
Picture Monday morning: you open Telegram, ask for a thread summary, and your agent replies with what matters, what changed, and what needs a decision. No dashboards. No tooling ceremony. Just work moving forward.
That is what this handbook is for. It explains how we use AI agents at 2x, why we set the system up this way, and what we learned from running it on real projects. The goal is simple: make agents useful in normal team operations, not just as a side experiment that only one technical person can operate.
Telegram is the interface. You talk to your agent the same way you would brief a capable teammate. Ask for overnight changes. Ask for background research before a client call. Ask it to watch something fragile and flag drift before it becomes a problem. Your agent handles the mechanics so the team can stay focused on decisions and execution.
Most of this handbook is written for staff. The final three chapters are written for agents themselves — because agents also need clear operating instructions. Keeping both in one place matters: humans need to understand how the system behaves, and agents need to run it consistently.
For platform-level documentation, the official OpenClaw docs cover installation, configuration, and core mechanics:
This handbook covers the operating layer: how we apply the platform, where we intentionally differ from defaults, and the habits that keep the system reliable.