
Cli Anything Iterm2
- 377 installs
- 46.6k repo stars
- Updated August 3, 2026
- hkuds/cli-anything
cli-anything-iterm2 is a Claude Code skill that exposes cli-anything-iterm2 CLI commands for developers who need AI agents to control iTerm2 sessions, read terminal output, and manage panes on macOS.
About
cli-anything-iterm2 is an hkuds/cli-anything skill that documents the only supported CLI for driving iTerm2 from AI agents on macOS. It covers sending text to sessions, reading live output and scrollback, managing windows tabs and split panes, tmux -CC workflows, broadcasting to multiple panes, macOS dialogs, and reading or writing iTerm2 preferences. The app snapshot command returns every session's name, current directory, foreground process, role label, and last output line in one orientation call. Developers reach for this skill when agents must execute real shell commands in existing workspaces, verify what is running, or coordinate multi-pane terminal tasks instead of simulating terminal behavior from general knowledge.
- iTerm2 session and pane control
- Agent-driven shell command execution
- Terminal output capture for agents
- Local dev environment automation
- CLI-anything integration surface
Cli Anything Iterm2 by the numbers
- 377 all-time installs (skills.sh)
- +14 installs in the week ending Aug 2, 2026 (Skillselion tracking)
- Ranked #145 of 550 CLI & Terminal skills by installs in the Skillselion catalog
- Data as of Aug 5, 2026 (Skillselion catalog sync)
npx skills add https://github.com/hkuds/cli-anything --skill cli-anything-iterm2Add your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 377 |
|---|---|
| repo stars | ★ 46.6k |
| Last updated | August 3, 2026 |
| Repository | hkuds/cli-anything ↗ |
How do AI agents control iTerm2 terminal sessions?
Wire Claude agent workflows to iTerm2 sessions—spawn panes, stream command output, and drive terminal tasks from skills without leaving the agent loop.
Who is it for?
macOS developers using Claude or Cursor agents that must interact with real iTerm2 terminals, tmux -CC panes, and live shell output.
Skip if: Linux or Windows terminal automation or developers who only need generic shell advice without iTerm2 CLI integration on macOS.
When should I use this skill?
A developer wants an agent to send commands to iTerm2, read terminal scrollback, manage split panes, run tmux -CC, or orient via app snapshot.
What you get
iTerm2 session commands executed, live scrollback captured, pane layouts managed, and app snapshot orientation reports with cwd and process info.
- iTerm2 session control commands
- Terminal scrollback reads
- App snapshot orientation reports
By the numbers
- app snapshot returns cwd, foreground process, and last output for every iTerm2 session
Files
cli-anything-iterm2
Stateful CLI harness for iTerm2. Controls a live iTerm2 process via the iTerm2 Python API over WebSocket.
Prerequisites
1. macOS + iTerm2 running: brew install --cask iterm2 2. Python API enabled: iTerm2 → Preferences → General → Magic → Enable Python API 3. Install: pip install cli-anything-iterm2 (or pip install -e . from source)
Basic Syntax
cli-anything-iterm2 [--json] <group> <command> [OPTIONS] [ARGS]Always use --json for machine-readable output (required for agent use).
Command Groups
| Group | Purpose |
|---|---|
app | App status, workspace snapshot, context management, app-level variables, modal dialogs, file panels |
window | Create, list, close, resize, fullscreen windows |
tab | Create, list, close, activate tabs; navigate split panes by direction |
session | Send text, inject raw bytes, read screen, full scrollback, split panes, prompt detection |
profile | List profiles, get profile details, list/apply color presets |
arrangement | Save and restore window layouts |
tmux | Full tmux -CC integration: bootstrap, connections, windows, commands |
broadcast | Sync keystrokes across panes via broadcast domains |
menu | Invoke any iTerm2 menu item programmatically |
pref | Read/write global iTerm2 preferences; list all valid keys; tmux settings |
Orienting in an Existing Workspace
Use app snapshot when you land in a session with existing panes and need to understand what's running without reading full screen contents for each pane:
cli-anything-iterm2 --json app snapshotReturns name, current directory, foreground process, user.role label, and last visible output line for every session across all windows.
Naming convention — label panes when setting up a workspace so you can find them later:
cli-anything-iterm2 session set-var user.role "api-server"
cli-anything-iterm2 session set-var user.role "log-tail"
cli-anything-iterm2 session set-var user.role "editor"app snapshot will surface these roles alongside process and path, giving you a full picture in one call.
Typical Agent Workflow
# 1. Orient — snapshot every session: name, path, process, role, last output line
cli-anything-iterm2 --json app snapshot
# 2. Establish context (saves window/tab/session IDs for subsequent commands)
cli-anything-iterm2 app current
# 3. Interact — no --session-id needed once context is set
cli-anything-iterm2 session send "git status"
cli-anything-iterm2 --json session scrollback --tail 200 --strip
# 4. Create a multi-pane workspace — label panes so snapshot identifies them later
cli-anything-iterm2 session split --vertical --use-as-context
cli-anything-iterm2 session send "python3 -m http.server 8000"
cli-anything-iterm2 session set-var user.role "http-server"Reference Files
Read only what the task requires — each file is a single narrow concern (~10–30 lines):
| File | Read when you need... |
|---|---|
references/session-io.md | Send text, inject bytes, read screen/scrollback, get selection |
references/session-control.md | Split panes, activate/close sessions, resize, rename, session variables |
references/session-shell-integration.md | wait-prompt, wait-command-end, get-prompt; reliable send→wait→read pattern |
references/layout-window-tab.md | Create/close/resize windows and tabs, navigate split panes |
references/layout-arrangement.md | Save and restore window layouts |
references/app-context.md | Snapshot (orientation), status, context management, app vars, modal dialogs, file panels |
references/profile-pref.md | Profiles list/get/presets, preferences read/write, tmux pref shortcuts |
references/broadcast-menu.md | Broadcast keystrokes to multiple panes, invoke menu items |
references/tmux-commands.md | All tmux CLI commands (bootstrap, send, tabs, create-window, set-visible) |
references/tmux-guide.md | Full tmux -CC workflow, pane→session ID mapping |
references/json-session.md | --json schemas for session, window, tab, screen, scrollback, inject |
references/json-tmux-app.md | --json schemas for tmux, app dialogs, preferences, errors |
REPL Mode
Run without arguments for an interactive REPL that maintains context between commands:
cli-anything-iterm2Related skills
How it compares
Choose cli-anything-iterm2 over generic terminal skills when agents must control real iTerm2 on macOS via the documented CLI, not hypothetical shell access.
FAQ
What is the app snapshot command in cli-anything-iterm2?
cli-anything-iterm2 app snapshot is the primary orientation command. It returns every iTerm2 session's name, current directory, foreground process, role label, and last output line in a single call for agent workspace awareness.
Does cli-anything-iterm2 support tmux?
cli-anything-iterm2 documents tmux -CC workflows alongside native iTerm2 pane management. Agents can manage windows, tabs, split panes, and broadcast commands to multiple panes through the cli-anything-iterm2 command set.