Claw Mux logo

Claw Mux

Community
LeeJuOh
claw-mux

Control cmux terminal topology and I/O — send commands to panes, read output, split layouts, monitor logs, orchestrate multi-pane workflows. Requires cmux environment.

Overview

PublisherLeeJuOh
Repositoryclaude-code-zero
Skill nameclaw-mux
Stars
53
Forks
4
Bundled files
8
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 8 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by LeeJuOh on GitHub. Read the source before you install it.

Installation

Install the Claw Mux AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/LeeJuOh/claude-code-zero/tree/main/plugins/claw-mux/skills/claw-mux
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/LeeJuOh/claude-code-zero.git /tmp/claude-code-zero
mkdir -p .claude/skills
cp -r /tmp/claude-code-zero/plugins/claw-mux/skills/claw-mux .claude/skills/claw-mux
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Claw Mux in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Claw Mux on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Claw Mux is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

cmux Core Control

Use this skill to control cmux topology, terminal I/O (send commands to other panes and read output), notifications, and sidebar metadata.

Environment Detection

cmux sets $CMUX_WORKSPACE_ID and $CMUX_SURFACE_ID in every terminal it spawns. Check before using cmux commands:

bash
if [ -z "$CMUX_WORKSPACE_ID" ]; then
  echo "Not running inside cmux — cmux commands unavailable"
fi

If not in cmux, fall back to standard tools (no pane splitting, no cmux notifications).

Core Concepts

  • Window: Top-level macOS cmux window.
  • Workspace: Tab-like group within a window (sidebar entry).
  • Pane: Split container in a workspace.
  • Surface: A tab within a pane (terminal or browser panel).

Hierarchy: Window > Workspace > Pane > Surface

Default output uses short refs: window:N, workspace:N, pane:N, surface:N. UUIDs are still accepted as inputs. Request UUID output only when needed: --id-format uuids|both.

Fast Start

Topology

bash
# Always identify first — short refs (pane:1, surface:N) are session-scoped
cmux identify --json                                          # current caller context
cmux list-workspaces                                          # list workspaces
cmux list-panes                                               # list panes in workspace
cmux list-pane-surfaces --pane pane:1                          # list surfaces in pane
cmux new-workspace                                            # create workspace
cmux new-split right                                           # split current surface to the right
cmux move-surface --surface surface:7 --pane pane:2            # move surface between panes
cmux trigger-flash --surface surface:7                         # visual flash

Terminal I/O (send commands to other panes, read output)

bash
cmux send --surface surface:N "echo hello\n"                   # send text + Enter (\n = Enter)
cmux send --surface surface:N "npm start\n"                    # run a command in another pane
cmux send-key --surface surface:N enter                        # send Enter key separately
cmux send-key --surface surface:N ctrl+c                       # send Ctrl+C (stop a process)
cmux read-screen --surface surface:N                           # read visible terminal text
cmux read-screen --surface surface:N --scrollback              # include scrollback history
cmux read-screen --surface surface:N --scrollback --lines 200  # last 200 lines only

Notifications & Sidebar

bash
cmux notify --title "Build Complete" --body "All tests passed"
cmux set-status build "Ready" --icon checkmark --color "#34c759"
cmux set-progress 0.75 --label "Running tests"
cmux log --level info --source claude-code "Starting build..."

Terminal I/O Patterns

The core workflow: split a pane, send a command, detect completion, read the output.

Before sending any command, decide how you'll know when it finishes. Claude Code blocks foreground sleep over 2 seconds — pick a completion strategy based on what you're running:

Command typeExampleStrategy
Batch (start→finish)npm run build, pytestwait-for signal
Long-running servicenpm run dev, docker compose upContent poll for "ready" text
Interactive programclaude, node, psqlContent poll for prompt marker
Instantecho, ls, catImmediate read-screen

Batch commands → wait-for

Chain cmux wait-for -S <signal> to the command. Run the wait with run_in_background: true.

bash
cmux new-split right  # → surface:9
cmux send --surface surface:9 "npm run build && cmux wait-for -S build-done\n"

# run_in_background: true — notified on completion
cmux wait-for build-done --timeout 120 && cmux read-screen --surface surface:9 --scrollback --lines 50

Services / interactive programs → content poll

Poll read-screen for expected text. Use the bundled helper ${CLAUDE_SKILL_DIR}/scripts/poll-screen.sh with run_in_background: true (references shorten it to poll-screen.sh):

bash
# run_in_background: true
${CLAUDE_SKILL_DIR}/scripts/poll-screen.sh surface:9 "ready|listening|Claude Code" --timeout 60

Common match patterns:

  • Server ready: ready|listening on|started
  • Claude Code ready: Claude Code|tips:|❯
  • Task complete: ╭─|❯|\\$\\s*$
  • Build result: ✓|passed|error|failed

Hybrid example

Server start → wait for ready → then batch work:

bash
cmux send --surface surface:9 "npm run dev\n"
# run_in_background: true — poll for server ready
${CLAUDE_SKILL_DIR}/scripts/poll-screen.sh surface:9 "ready on" --timeout 30

# (after background notification confirms ready)
cmux send --surface surface:9 "curl localhost:3000/health && cmux wait-for -S health-ok\n"
cmux wait-for health-ok --timeout 10

For detailed patterns (server monitoring, E2E testing, build pipelines, remote Claude Code), read the terminal-io reference.

Sidebar Metadata

bash
# status pill: set-status <key> <value> [--icon name] [--color #hex]
cmux set-status build "Ready" --icon checkmark --color "#34c759"
cmux set-status build "Failed" --icon xmark --color "#ff3b30"
cmux clear-status build

# progress bar (0.0 to 1.0, optional --label)
cmux set-progress 0.75 --label "Running tests"
cmux clear-progress

# log messages (leveled, optional --source)
cmux log --level info --source claude-code "Starting build..."
cmux log --level success "Build complete"
cmux log --level warning "3 deprecation warnings"
cmux log --level error "Test suite failed"

Coexistence

  • In cmux: Use cmux new-split for panes, cmux send/cmux read-screen for terminal I/O, cmux notify for notifications.
  • Outside cmux: These features are unavailable. Use standard terminal workflows.
  • Browser automation in cmux: Use the cmux-browser skill for embedded webview automation.
  • Markdown viewing in cmux: Use cmux markdown open <path> directly to open a live-reloading panel.
  • With cmux claude-teams: Teams uses its own tmux-shim layer for topology — do not mix direct cmux topology commands with Teams orchestration.

Gotchas

  • cmux commands fail silently outside cmux — always check $CMUX_WORKSPACE_ID before using any cmux command.
  • Short refs (surface:7) are session-scoped and may change between sessions — always cmux identify --json at the start of automation to discover current topology.
  • send includes \n for Enter — cmux send "command\n" sends the command and presses Enter in one call. No need for a separate send-key enter unless you want explicit timing control.
  • send-keys does not exist — the command is send (text) and send-key (key event). This is different from tmux's send-keys.
  • press does not exist — to send Ctrl+C, use cmux send-key --surface surface:N ctrl+c.
  • read-screen returns the current screen state — if a command is still running, the output may be incomplete. Use --scrollback --lines N for history, or use a completion strategy (see Terminal I/O Patterns above).
  • In Claude Code, sleep over 2 seconds is blocked in foreground — use wait-for for batch commands, content polling for interactive programs. Both require run_in_background: true. Avoid sleep N && read-screen — it reads too early or too late.
  • Always plan completion detection before sending a command — decide upfront: is this batch (wait-for), service/interactive (content poll), or instant (immediate read)? Sending without a plan leads to blind retries.
  • set-status --color requires hex values ("#34c759") — named colors like green are not supported.
  • set-status requires a key (e.g., build) as the first argument — the key allows multiple independent status pills per workspace.
  • If cmux claude-teams is active, it manages topology through a tmux-compat shim. Direct topology commands (new-split, move-surface) may conflict with Teams orchestration.
  • Desktop alerts are suppressed when the cmux window is focused and the workspace is active — use cmux log for always-visible status updates.

Deep-Dive References

ReferenceWhen to Use
Terminal I/OSend commands, read output, server monitoring, E2E testing, build pipelines
Sync and Automationwait-for synchronization, buffers, cmux.json custom workspace commands
Handles and IdentifyHandle syntax, self-identify, caller targeting
Windows and WorkspacesWindow/workspace lifecycle and reorder/move
Panes and SurfacesSplits, surfaces, move/reorder, focus routing
Trigger Flash and HealthVisual flash confirmation and surface health checks
NotificationsNotification CLI, sidebar metadata API, hook patterns

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Claw Mux AI skill do?

Control cmux terminal topology and I/O — send commands to panes, read output, split layouts, monitor logs, orchestrate multi-pane workflows. Requires cmux environment.

Why use Claw Mux on TypingMind?

Because you install it once and use it with any model. Claw Mux is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Claw Mux in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/LeeJuOh/claude-code-zero/tree/main/plugins/claw-mux/skills/claw-mux. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Claw Mux?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Claw Mux?

As many as you like. As long as a model supports skills, you can use Claw Mux with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Claw Mux AI skill free?

Yes. It is published on GitHub by LeeJuOh under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇