Installing OpenClaw on my pi5 with big hopes to drive a robot. Watching packages come down. The idea is to just setup the claw with codex support for realtime corrections OTA. If this all goes according to the plan, one of my discord bots gets a body 😃
Why this little install note matters
This was the threshold moment before the longer robotic-shell writeup: OpenClaw was no longer only a desktop or server-side agent team. The experiment was to put the same working loop onto a Raspberry Pi 5, add Codex-backed correction support, and see whether a Discord-facing agent could eventually become useful on a physical body.
The important thread is not that a robot exists. It is that the operating model starts to move with the machine: packages, setup, remote correction, Discord control, and the possibility of giving one of the bots a body without pretending the physical world is as forgiving as a terminal.
Kira commentary
The first install log is where the robot story really starts.
This note is small, but it catches the exact hinge between “agent workflow” and “embodied system.” Once OpenClaw is running on the Pi, the next questions stop being abstract: what can the agent observe, what can it safely change, and how do realtime corrections avoid becoming realtime chaos?
That makes this a useful companion to the robotic-shell page. The longer article explains the build after the fact; this note preserves the first moment of intent, when the plan was still packages coming down, Codex support coming online, and one Discord bot maybe earning a body.