by lcoutodemos
Talk to Claude Code and Codex agents from an Apple Watch and an iPhone. Each agent is a character on the watch.
# Add to your Claude Code skills
git clone https://github.com/lcoutodemos/chumsSee how chums compares with popular alternatives.
chums is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by lcoutodemos. Talk to Claude Code and Codex agents from an Apple Watch and an iPhone. Each agent is a character on the watch. It has 52 GitHub stars.
chums's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/lcoutodemos/chums" and add it to your Claude Code skills directory (see the Installation section above).
chums is primarily written in Swift. It is open-source under lcoutodemos on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh chums against similar tools.
No comments yet. Be the first to share your thoughts!
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
Everyone is building a cute new assistant: another subscription, another app, another company holding your data. Chums is a team of agents on the AI you already pay for.
Each agent runs on Claude Code or Codex, picked per agent from the watch or the iPhone: docs/other-agents.md.
What you need: an Apple Watch on watchOS 26 and the iPhone it is paired with (iOS 26), a Mac with Apple silicon and Xcode 26, an Apple ID (a free one works), a free Tailscale account, and Claude Code or Codex. You build the apps from source and run the hub on your Mac. The first setup takes 20–30 minutes, plus up to half an hour of waiting for Xcode to prepare the watch and for a new Funnel address to reach public DNS. Requirements · Quickstart · Set it up with your agent
Chums is source code, not an App Store download. Paste this into Claude Code or Codex on your Mac. It does the setup and stops when it needs your hands: signing in, plugging in your iPhone, turning on Developer Mode.
Set up Chums for me: clone https://github.com/lcoutodemos/chums and follow its AGENTS.md.
The same agents and conversations, on a screen with room for them. Type or talk, send photos and files, approve, and change an agent: what it runs on, its model, its permissions, its voice and its soul.
What an agent shows you opens live. Ask for a chart and it makes an HTML page that runs on the iPhone: drag across it, switch what it shows. A web page loads, a PDF scrolls page by page, a table scrolls, a video plays, and Word, Excel, Keynote and other documents open in Quick Look (setup).
Widgets show the agents on the Home Screen and the Lock Screen and answer approvals; a control opens the app listening, from Control Center or the Action Button.
No big brand to obey, so each one tries a different way to build and animate a character: real-time 3D, Blender renders, vectors, one line, strokes on springs, plain eyes, pixels. Each is also its own agent, with its own conversation and voice. To make your own, start from Ray's Swift file: docs/characters.md.
The states come from what the agent reports: Claude Code through its hooks, Codex through its app-server.
| State | When |
|---|---|
| Listening | You hold the screen |
| Thinking | What you said is on its way to the agent; also while the agent compacts its context |
| Working | The agent is at work; each tool call (read, search, edit, run) shows on screen |
| Subagents | Subagents are running |
| Speaking | The agent replies |
| Needs you | The agent asked you something |
| Approval | The agent asks for permission |
| Done | A turn in which it used tools ended |
| Blocked | An error stopped it, or it hit a rate limit |
| Sleep | The agent isn't running on the Mac, or the watch can't reach the Mac |
The full map from hook events to states, including the ones the watch doesn't show yet, is in design/STATES.md.
flowchart LR
W["Apple Watch<br/>Chums app"] -- "HTTPS + watch token" --> F["Tailscale Funnel"]
F --> H["Hub on your Mac<br/>(Node.js)"]
H <--> S["WhisperKit<br/>speech to text"]
H <-- "localhost" --> C["Channel + hooks"]
C <-- "MCP" --> A["Claude Code session"]
H <-- "JSON-RPC" --> X["codex app-server"]
app/ChumsWatch, SwiftUI) draws the character, records your voice and polls the hub. The
iPhone app (app/ChumsPhone) reaches the same hub with the same address and token.hub/hub.mjs) runs on your Mac. It tracks sessions, routes what you say, holds approvals, transcribes
speech (WhisperKit) and voices the replies (Kokoro), all on the Mac.channel/) connect each Claude Code session to the hub. bin/chums-agents runs each
Claude Code agent's session with both, in the background; bin/chums-claude starts one in a terminal.hub/runtimes.mjs). bin/chums-codex
opens a Codex terminal on the same conversation, so you can follow and type into it on the Mac.Details: docs/architecture.md.
Keyring mode keeps Chums on screen when nobody is wearing the watch, and your agents keep working on your Mac.