Tmux logo

Tmux

CommunityPopular
mitsuhiko
tmux

Remote control tmux sessions for interactive CLIs (python, gdb, etc.) by sending keystrokes and scraping pane output.

Overview

Publishermitsuhiko
Repositoryagent-stuff
Skill nametmux
Stars
3.1K
Forks
224
Bundled files
2
LicenseVibecoded
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.

  • 2 bundled files

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

  • Open source

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

Installation

Install the Tmux 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.
  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/mitsuhiko/agent-stuff.git /tmp/agent-stuff
mkdir -p .claude/skills
cp -r /tmp/agent-stuff/skills/tmux .claude/skills/tmux
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Tmux 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 Tmux 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 Tmux 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.

tmux Skill

Use tmux as a programmable terminal multiplexer for interactive work. Works on Linux and macOS with stock tmux; avoid custom config by using a private socket.

Quickstart (isolated socket)

bash
SOCKET_DIR=${TMPDIR:-/tmp}/claude-tmux-sockets  # well-known dir for all agent sockets
mkdir -p "$SOCKET_DIR"
SOCKET="$SOCKET_DIR/claude.sock"                # keep agent sessions separate from your personal tmux
SESSION=claude-python                           # slug-like names; avoid spaces
tmux -S "$SOCKET" new -d -s "$SESSION" -n shell
tmux -S "$SOCKET" send-keys -t "$SESSION":0.0 -- 'python3 -q' Enter
tmux -S "$SOCKET" capture-pane -p -J -t "$SESSION":0.0 -S -200  # watch output
tmux -S "$SOCKET" kill-session -t "$SESSION"                   # clean up

After starting a session ALWAYS tell the user how to monitor the session by giving them a command to copy paste:

To monitor this session yourself:
  tmux -S "$SOCKET" attach -t claude-lldb

Or to capture the output once:
  tmux -S "$SOCKET" capture-pane -p -J -t claude-lldb:0.0 -S -200

This must ALWAYS be printed right after a session was started and once again at the end of the tool loop. But the earlier you send it, the happier the user will be.

Socket convention

  • Agents MUST place tmux sockets under CLAUDE_TMUX_SOCKET_DIR (defaults to ${TMPDIR:-/tmp}/claude-tmux-sockets) and use tmux -S "$SOCKET" so we can enumerate/clean them. Create the dir first: mkdir -p "$CLAUDE_TMUX_SOCKET_DIR".
  • Default socket path to use unless you must isolate further: SOCKET="$CLAUDE_TMUX_SOCKET_DIR/claude.sock".

Targeting panes and naming

  • Target format: {session}:{window}.{pane}, defaults to :0.0 if omitted. Keep names short (e.g., claude-py, claude-gdb).
  • Use -S "$SOCKET" consistently to stay on the private socket path. If you need user config, drop -f /dev/null; otherwise -f /dev/null gives a clean config.
  • Inspect: tmux -S "$SOCKET" list-sessions, tmux -S "$SOCKET" list-panes -a.

Finding sessions

  • List sessions on your active socket with metadata: ./scripts/find-sessions.sh -S "$SOCKET"; add -q partial-name to filter.
  • Scan all sockets under the shared directory: ./scripts/find-sessions.sh --all (uses CLAUDE_TMUX_SOCKET_DIR or ${TMPDIR:-/tmp}/claude-tmux-sockets).

Sending input safely

  • Prefer literal sends to avoid shell splitting: tmux -L "$SOCKET" send-keys -t target -l -- "$cmd"
  • When composing inline commands, use single quotes or ANSI C quoting to avoid expansion: tmux ... send-keys -t target -- $'python3 -m http.server 8000'.
  • To send control keys: tmux ... send-keys -t target C-c, C-d, C-z, Escape, etc.

Watching output

  • Capture recent history (joined lines to avoid wrapping artifacts): tmux -L "$SOCKET" capture-pane -p -J -t target -S -200.
  • For continuous monitoring, poll with the helper script (below) instead of tmux wait-for (which does not watch pane output).
  • You can also temporarily attach to observe: tmux -L "$SOCKET" attach -t "$SESSION"; detach with Ctrl+b d.
  • When giving instructions to a user, explicitly print a copy/paste monitor command alongside the action don't assume they remembered the command.

Spawning Processes

Some special rules for processes:

  • when asked to debug, use lldb by default
  • when starting a python interactive shell, always set the PYTHON_BASIC_REPL=1 environment variable. This is very important as the non-basic console interferes with your send-keys.

Synchronizing / waiting for prompts

  • Use timed polling to avoid races with interactive tools. Example: wait for a Python prompt before sending code:
    bash
    ./scripts/wait-for-text.sh -t "$SESSION":0.0 -p '^>>>' -T 15 -l 4000
  • For long-running commands, poll for completion text ("Type quit to exit", "Program exited", etc.) before proceeding.

Interactive tool recipes

  • Python REPL: tmux ... send-keys -- 'python3 -q' Enter; wait for ^>>>; send code with -l; interrupt with C-c. Always with PYTHON_BASIC_REPL.
  • gdb: tmux ... send-keys -- 'gdb --quiet ./a.out' Enter; disable paging tmux ... send-keys -- 'set pagination off' Enter; break with C-c; issue bt, info locals, etc.; exit via quit then confirm y.
  • Other TTY apps (ipdb, psql, mysql, node, bash): same pattern—start the program, poll for its prompt, then send literal text and Enter.

Cleanup

  • Kill a session when done: tmux -S "$SOCKET" kill-session -t "$SESSION".
  • Kill all sessions on a socket: tmux -S "$SOCKET" list-sessions -F '#{session_name}' | xargs -r -n1 tmux -S "$SOCKET" kill-session -t.
  • Remove everything on the private socket: tmux -S "$SOCKET" kill-server.

Helper: wait-for-text.sh

./scripts/wait-for-text.sh polls a pane for a regex (or fixed string) with a timeout. Works on Linux/macOS with bash + tmux + grep.

bash
./scripts/wait-for-text.sh -t session:0.0 -p 'pattern' [-F] [-T 20] [-i 0.5] [-l 2000]
  • -t/--target pane target (required)
  • -p/--pattern regex to match (required); add -F for fixed string
  • -T timeout seconds (integer, default 15)
  • -i poll interval seconds (default 0.5)
  • -l history lines to search from the pane (integer, default 1000)
  • Exits 0 on first match, 1 on timeout. On failure prints the last captured text to stderr to aid debugging.

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 Tmux AI skill do?

Remote control tmux sessions for interactive CLIs (python, gdb, etc.) by sending keystrokes and scraping pane output.

Why use Tmux on TypingMind?

Because you install it once and use it with any model. Tmux 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 Tmux in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/mitsuhiko/agent-stuff/tree/main/skills/tmux. 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 Tmux?

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 Tmux?

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

Is the Tmux AI skill free?

Yes. It is published on GitHub by mitsuhiko under the Vibecoded 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 👇