copilot+

Drop-in wrapper for GitHub Copilot CLI: voice, screenshots, clipboard, a prompt queue, workflows, a safety scanner, and a live multi-session monitor. It does not replace Copilot CLI — you still need a Copilot subscription and copilot on your PATH.

npm install -g copilot-plus
copilot+

macOS Homebrew: brew tap Errr0rr404/copilot-plus && brew install copilot-plus

---

Hotkeys

Key	Action
? / Ctrl+/	Cheatsheet (from live config)
Ctrl+K	Command palette
Ctrl+R	Voice record (local whisper.cpp)
Ctrl+P	Screenshot → @/path.png
Ctrl+Y	Paste clipboard (text or image)
Ctrl+G	Git status + diff (optional recent files)
Ctrl+O	Attach a project file as @/abs/path
Ctrl+S	Stash the current draft
Ctrl+E	Open last response in $EDITOR
Ctrl+T	Toggle text-to-speech
Ctrl+B	Bookmark last response
Enter (while busy)	Queue the prompt until idle
!cmd	Run a shell command; inject stdout
Opt+Shift+1–4 / Ctrl+Shift+1–4	Workhorse model slots
Opt+Shift+5 / Ctrl+Shift+5	Toggle Auto Mode
Opt+1–9 / Ctrl+1–9	Macros

Everything else is forwarded to Copilot CLI as usual.

Commands

Command
copilot+	Launch with enhancements
copilot+ --monitor	Live multi-session dashboard
copilot+ --history [q]	Search prompts (--clear to wipe)
copilot+ --snippets	Bookmarks (--bookmarks alias)
copilot+ --usage	Local history + premium quota
copilot+ --workflows	List prompt chains
copilot+ --worktree [name]	Isolated git worktree
copilot+ --export-config / --import-config	Share team config
copilot+ --doctor	Validate config + environment
copilot+ --setup	Dependencies + microphone
copilot+ --preferences	Re-run onboarding
copilot+ --ask "q"	One-shot query
copilot+ --help-plus	Full help
copilot+ --version	Print version

---

Requirements

	macOS	Windows	Linux
Copilot CLI + Node ≥ 18	required	required	required
ffmpeg + whisper.cpp (voice)	brew install ffmpeg whisper-cpp	winget FFmpeg + whisper.cpp release	distro ffmpeg; build whisper.cpp
Clipboard / screenshot / TTS	built-in	built-in	wl-clipboard or xclip; grim+slurp or scrot; spd-say / espeak-ng

Download a Whisper model (not shipped on npm):

mkdir -p ~/.copilot/models
curl -L "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.en.bin" \
  -o ~/.copilot/models/ggml-base.en.bin

Then copilot+ --setup. For other languages, use a non-.en model and set voiceLanguage (ISO 639-1 or "auto").

---

Daily use

Voice. Ctrl+R to start, speak, Ctrl+R to stop. Transcription is local. Ctrl+C cancels. Optional wake phrase (wakeWord.enabled) listens for “hey copilot”.

Screenshots. Ctrl+P opens the OS region picker and injects @/path. Paths with spaces are quoted.

Clipboard / git / shell. Ctrl+Y pastes text or an image. Ctrl+G injects git status and diff (contextHotkey.kinds can add "recent" / "branch"). !ls -la runs in $SHELL (15s / 16 KiB cap) and injects a fenced block.

Safety. Outgoing prompts are scanned for common keys and private-key headers. Choose send / redact / cancel, or set safety.autoRedact: true. Disable with safety.enabled: false.

Queue & stash. While Copilot is answering, Enter queues the line. Ctrl+S parks a draft (~/.copilot/prompt-stash.json).

Workflows. Drop YAML/JSON in ~/.copilot/workflows/ and run from the palette. Variables: {{cwd}}, {{git_branch}}, {{clipboard}}, {{date}}.

name: Fix & Test
steps:
  - prompt: "Run the failing tests and show the errors"
    wait: true
  - prompt: "Fix the failures"
    wait: true

Monitor. copilot+ --monitor — j/k select, / filter, s sort, K SIGTERM, q quit.

Plugins. Put *.js / *.cjs in ~/.copilot/plugins/. Hooks: beforeSend, afterReceive, afterPrompt, onModelSwitch, onBookmark. Reload from the palette. Errors in one plugin are isolated.

---

Configuration

~/.copilot/copilot-plus.json (created on first run). Re-run copilot+ --preferences anytime.

Key	Default
modelPath	auto-detected	Whisper .bin
audioDevice	auto-detected	ffmpeg input (":2" / device name / pulse:default)
autoSubmit	false	Enter after voice
voicePreview	false	Edit transcription before send
voiceLanguage	"en"	en / auto / ISO 639-1
theme	"dark"	dark light solarized monokai auto
shellPrefix	"!"	Shell-exec prefix
historyEnabled	true	~/.copilot/history.jsonl
workhorseModels	empty	Slots 1–4
autoModels	empty	fast / medium / powerful (fallback: slots 1 / 1 / 2)
macros	empty	Slots 1–9
wakeWord	off	{ enabled, phrase, chunkSeconds }
tts	off	{ enabled, rate, voice }
safety	on	{ enabled, disabledKinds, autoRedact }
contextHotkey.kinds	["status","diff"]	also recent, branch
notifications	idle / waiting / quota	quiet, webhookUrl, rules (os / bell / webhook)
queue	{ whenBusy: true, maxSize: 20 }
usage	{ dailyBudget: 0, handoffAtPct: 90 }
workflowsDir	~/.copilot/workflows

Quota lookup uses COPILOT_GITHUB_TOKEN, then GH_TOKEN / GITHUB_TOKEN, then gh auth token. Log level: COPILOT_PLUS_LOG. Theme override: COPILOT_PLUS_THEME.

Config export copies settings and workflows only — not GitHub tokens or notification webhook URLs.

---

Troubleshooting

posix_spawnp failed — reinstall: npm install -g copilot-plus (postinstall fixes node-pty helper permissions).

Wrong mic / same word every time — copilot+ --setup, or set audioDevice from ffmpeg -f avfoundation -list_devices true -i "" (macOS) / ffmpeg -f dshow -list_devices true -i dummy (Windows) / pactl list short sources (Linux, pulse:<name>).

Mic permission — enable the terminal app under OS privacy settings.

No whisper model — download ggml-base.en.bin as above, then --setup.

Wake word silent — shorter phrase; raise wakeWord.chunkSeconds to 3–4.

Opt+Shift / Opt+N dead in Apple Terminal — Settings → Profiles → Keyboard → Use Option as Meta Key. kitty/WezTerm/Windows Terminal: send CSI u (WezTerm: enable_kitty_keyboard = true).

Screenshot / clipboard / TTS no-op on Linux — install grim+slurp or scrot; wl-clipboard or xclip; spd-say or espeak-ng.

Safety false positive — r redact, or safety.disabledKinds: ["github_token"].

copilot+ --doctor checks binaries, model, and config in one pass.

---

License

MIT © Errr0rr404

Changelog · Roadmap · Issues