BuildBB-0132

code-chan: coding-agent status and permission prompts on an M5Stack CoreS3

code-chan runs on a Stack-chan robot with an M5Stack CoreS3 and shows whether Claude Code or Codex is ready, working or waiting for permission. The user approves or rejects a prompt on its touchscreen. A local Node.js bridge sends hook events over USB serial, and the firmware checks each message before acting. README photos show three states.

Open original source github.com · stelis-dev

Shown working
Why this evidence label?

README shows photos of the device in ready, working and permission-prompt states.

How we check entries
Preview image for code-chan: coding-agent status and permission prompts on an M5Stack CoreS3
The source's own preview image, from github.com.

Briefing

The bridge follows one session at a time: the first prompt sent while the device is ready takes it, and other sessions keep their normal permission prompts. Each serial connection is numbered, and the firmware accepts messages only from the current one. A permission screen shows only the tool's name, at most 32 bytes, never the command or file path. An answer counts only through the labeled APPROVE or REJECT control, and only while the waiting hook, session, task, connection and request ID all still match.

Installation is from source and needs PlatformIO, Python 3, Node.js 22 and a macOS or Linux host. On Claude Code's CLI an optional status line puts remaining context and usage in the caption. The firmware redraws at 24 frames per second; the README says whether that holds on hardware is a separate check.

From the GitHub API

Repository facts

Repository facts for stelis-dev/code-chan, from the GitHub API
Repositorystelis-dev/code-chan
Stars2
LicenseApache-2.0 Permissive
LanguageC++
Last push6 Aug 2026 2 mo ago

Tags

Board
Interface

Report a problem with this entry