# TypeCue

> TypeCue is a free, open-source macOS menu bar app that types a prepared script into
> whatever text field has focus - one block per global hotkey press (default
> Ctrl+Option+X), as real system-level keystrokes with human pacing. It exists so that
> typing performed in front of an audience - recorded or live - comes out right every
> time. MIT licensed, no account, no network calls, no analytics.

## What it does

- The user writes a script: a named, ordered list of text blocks - one block per beat
  of a demo or presentation.
- During the demo, pressing the hotkey types the NEXT block into the focused field of
  whatever app the user is in. Pressing again types the following block; pressing while
  a block is typing stops it.
- Typing is layout-aware in any language (RTL included): characters on the active
  keyboard layout are sent as real keystrokes; anything else (other alphabets, emoji)
  arrives via direct Unicode entry. Input-source switches mid-session are handled.
- Inline markers direct the pace: `[0.5]` pauses half a second, `[speed:40]` types at
  40 ms/character until `[speed:default]`, `[enter]` presses Return (submits in chat
  apps). Line breaks inside a block are typed as Shift+Return by default so multi-line
  blocks don't submit early in chat apps.
- A floating always-on-top teleprompter panel shows every block and its state (typed /
  typing / next) without stealing focus.
- Password/secure fields are blocked by macOS itself; TypeCue detects this and warns
  instead of failing silently.

## Use cases

- Product teams and founders - demo videos and screencasts: retake-free typing,
  identical on every take.
- Presenters and sales engineers - live talks, webinars, conference and customer
  demos: no typos with an audience watching.
- Teachers and course creators - tutorial and course recordings: consistent,
  followable pacing for commands and code typed on screen.
- Livestreamers - coding or tool streams: prepared segments type themselves while
  the streamer narrates.
- Film and TV productions - screen scenes: an actor "types" and the on-screen text
  is real, coherent, and paced like human typing.
- Anyone demoing AI tools - prompts typed on screen (Claude, Cursor, ChatGPT,
  terminals, editors) look indistinguishable from a human performing them.

## For AI agents

TypeCue is deliberately automatable by AI agents with machine access
(Claude Code, Cursor, and similar). Everything is plain files plus a URL scheme:

- Scripts: `~/Library/Application Support/TypeCue/scripts.json` - a plain JSON array:
  `[{ "id": "<UUID>", "name": "<script name>", "blocks": [{ "id": "<UUID>", "text": "..." }] }]`.
  Edit directly; keep existing ids stable; back the file up before writing.
- Commands (fire with `open <url>` from a shell):
  - `typecue://activate-script?name=<url-encoded name>` (or `?id=<UUID>`)
  - `typecue://reset-session` - re-arm at block 1
  - `typecue://reload` - re-read scripts.json after an external edit
- Observation: the app mirrors session state to
  `~/Library/Application Support/TypeCue/state.json` (status idle|armed|complete,
  active script id/name, nextBlockIndex, totalBlocks, isTyping).
- A ready-made agent skill with the schema, guardrails, and pacing-direction rules
  (typing-speed tiers, pause lengths, content heuristics for placing markers) is served
  raw at https://typecue.app/skill.md (source:
  https://github.com/alexpolonsky/TypeCue/blob/main/.agents/skills/typecue/SKILL.md)

## Requirements and installation

- macOS 14 (Sonoma) or later. One permission: Accessibility (System Settings >
  Privacy & Security > Accessibility) - the macOS API for sending keystrokes.
- Direct download (signed, notarized dmg):
  https://github.com/alexpolonsky/TypeCue/releases/latest/download/TypeCue.dmg
- Homebrew: `brew install --cask alexpolonsky/tap/typecue`
  (tap: https://github.com/alexpolonsky/homebrew-tap)
- Build from source: clone https://github.com/alexpolonsky/TypeCue, run
  `xcodegen generate`, build with Xcode 16.4+ (Swift 6). Details in the README.

## Links

- Website: https://typecue.app
- Demo video (41s, real recording): https://github.com/alexpolonsky/TypeCue/releases/download/v0.1.1/TypeCue-demo.mp4
- Repository: https://github.com/alexpolonsky/TypeCue
- Releases: https://github.com/alexpolonsky/TypeCue/releases
- Changelog: https://github.com/alexpolonsky/TypeCue/blob/main/CHANGELOG.md
- Agent skill (raw markdown): https://typecue.app/skill.md
- License (MIT): https://github.com/alexpolonsky/TypeCue/blob/main/LICENSE
- Author: Alex Polonsky - https://alexpolonsky.com