Command line

Everything you do in Azynote with a click or a shortcut, a script can do too. azynote-ctl sends one command to the Azynote that is running on your Mac, prints the answer and exits:

azynote-ctl transcription start --session 20261001110000
azynote-ctl transcription stop
azynote-ctl onepager generate --session 20261001110000

Use it from Terminal, from a hook, from a Shortcuts "Run Shell Script" action, or from any key or launcher that can run a command.

Set it up#

Azynote writes the command for you every time it starts, at ~/.azynote/bin/azynote-ctl. To type it as azynote-ctl from anywhere, add that folder to your shell's path once:

echo 'export PATH="$HOME/.azynote/bin:$PATH"' >> ~/.zshrc

Open a new Terminal window and try azynote-ctl status. Azynote has to be running: when it isn't, azynote-ctl says so and does nothing.

Hooks don't need this step. Every hook gets the full path as $AZYNOTE_CTL (see below).

Find your way around#

• azynote-ctl commands lists every command, with one line on what it does.
• azynote-ctl help transcription start shows one command's flags.
• azynote-ctl status shows what Azynote is doing right now: the transcription, dictation, the voice conversation, the window, the open session.

Flags are words, never JSON: --title "Weekly sync". A flag that takes several values is repeated: --subject-keys acme --subject-keys beta. A yes-or-no flag is on by itself and off with no- in front: --include-content or --no-include-content.

Control a transcription#

azynote-ctl transcription start --session 20261001110000
azynote-ctl transcription start --new --subject-keys acme --title "Kickoff"
azynote-ctl transcription pause
azynote-ctl transcription resume
azynote-ctl transcription mic pause
azynote-ctl transcription stop

start always needs to know which session: an existing one with --session, or a new one with --new (a new session belongs to a subject, so give its key). Azynote never guesses from what happens to be on screen. The session opens in the window, so you see it being transcribed.

Asking for what is already true is fine: stop when nothing is transcribing, or pause when already paused, succeeds and changes nothing. Asking for something Azynote wouldn't let you do with a click, like starting a second transcription while one is running, is refused with the reason.

transcription toggle is made for a single key: it pauses a live transcription and resumes a paused one. It never starts or stops one, so pressing it by accident can't end your meeting.

Stopping a transcription doesn't write the OnePager. onepager generate --session <id> does, the same way the sparkle button does, with the templates chosen for that session.

sync start runs Smart Sync on the open workspace, as Tools > Smart Sync does: every missing or outdated OnePager and one-liner, then the status of their subjects. It works in the background, so you can run it from a script at the end of the day. Add --show to watch it in its window. Closing that window doesn't stop the sync, and wait sync=completed --timeout 1800 waits until it's finished.

Wait for something to happen#

wait returns as soon as a value in status is the one you name, so a script never has to guess how long to sleep:

azynote-ctl transcription stop
azynote-ctl wait transcription=idle --timeout 600
azynote-ctl onepager generate --session 20261001110000
azynote-ctl wait onepager=completed --session 20261001110000 --timeout 300

It gives up after --timeout seconds (30 unless you say otherwise) and tells you what the value is instead.

Dictation, voice and the window#

azynote-ctl dictation toggle      # the same as pressing ⌃⇧D
azynote-ctl voice toggle          # start, then send: like ⌃⇧V
azynote-ctl window mini
azynote-ctl window pin on
azynote-ctl open settings --tab hooks
azynote-ctl open "azynote://workspaces/…/sessions/20261001110000"

Dictation types into the app that is in front when it starts, exactly like the shortcut. The voice commands drive the turn-by-turn conversation; when Voice is set to a live engine, they say so and leave it to the chat window. open takes the same azynote:// links to a session, a OnePager or a note that the chat gives you.

Ask the orb, and show what you want on screen#

azynote-ctl orb ask "What did Acme ask for on today's call?"
azynote-ctl wait orb=answered --timeout 180
azynote-ctl orb follow-up "Draft a short email to Karen with our next steps."
azynote-ctl wait orb=answered --timeout 180
azynote-ctl orb chat

orb ask opens the orb's card over the app in front, types the question in letter by letter and asks it, as a person would. The card opens at the mouse pointer, or at --x and --y (screen points) when you give them. The command returns once the question is asked, so wait orb=answered is how a script waits for the answer.

• orb follow-up types a follow-up under the answer and asks it in the same thread. It works only while the card shows an answer.
• orb chat does what the chat button on the card does: the conversation continues in the chat window, on its thread. wait chat=open waits for the window.
• orb close closes the card, like its ×. An answer still coming is stopped.
• orb type types into the card without asking, so you can see what it offers for the text, like the list of actions after /.
• orb tap taps the orb itself: it opens what waits behind it (a meeting, the agent's notices, your catch-up after /brb), or the ask card.
• view calendar, view timeline, view tasks or view subjects shows one of the main views, like Ctrl+1 to Ctrl+4.
• session tab note, session tab transcript or session tab onepager shows that tab of the open session, like clicking it. During a transcription, transcript shows the live one.

As everywhere on this page, a command Azynote can't carry out right now is refused with the reason: a follow-up with no answer on the card, or a session tab with no session open. Together they make scripted demos and walkthroughs easy.

Set up a workspace's AI engine#

azynote-ctl ai preset acp
azynote-ctl acp command "goose acp"
azynote-ctl acp allow-all on

• ai preset gemini, openai, local, acp or cli routes every AI capability of the workspace to one engine, the same as the preset in Settings, AI engines. If the embedding engine changes, the command says so: workspace search needs a reindex from Settings, Search, as Settings warns too.
• acp command sets the command that launches your ACP agent in this workspace. It tells you if Azynote doesn't find it on your PATH.
• acp allow-all on or off is Allow every tool in Settings, ACP (see Chat). With it on, the agent runs its tools without asking, its own shell included.

All three change only the workspace open in the window.

Use it from a hook#

Every hook gets $AZYNOTE_CTL, the command of the Azynote that ran the hook. Two ready-made examples on the Hooks page use it:

Transcribe a meeting you prepared, as soon as it starts (Meeting starts)

case "$AZYNOTE_MEETING_RESPONSE" in
  accepted|organizer) ;;
  *) exit 0 ;;
esac
if [ -n "$AZYNOTE_SESSION_ID" ] && [ "$AZYNOTE_TRANSCRIBING" != yes ]; then
  "$AZYNOTE_CTL" transcription start --session "$AZYNOTE_SESSION_ID"
fi

$AZYNOTE_SESSION_ID is the meeting's session when you already opened the meeting from the calendar, empty when you haven't, so meetings you didn't prepare are left alone.

Write the OnePager when the transcription is done (Transcription stops)

"$AZYNOTE_CTL" wait transcription=idle --timeout 600 && \
  "$AZYNOTE_CTL" onepager generate --session "$AZYNOTE_SESSION_ID"

The wait lets the last few seconds of speech finish transcribing before the OnePager is written.

Use it in a script#

Add --json to get the answer as JSON instead of text, with the state Azynote is in afterwards under status:

azynote-ctl --json status | jq -r '.status.transcription'

The exit code tells your script what happened:

Good to know#

• It talks to your Azynote only. The command connects to Azynote on your own Mac, as you. Nothing goes over the network.
• Changes show up in the app right away. A subject or a note created from the command line appears in the window, like one created by an AI tool through the MCP server.
• Data commands work on the workspace open in the window. workspace open <folder> switches it, the same way the workspace picker does, and is refused while you are transcribing. The engine and permission commands (ai preset, acp ...) work on that workspace too, and only on it.
• Turning the mic on is for the command line only. The commands that start a transcription, dictation or a voice conversation are not offered to AI tools connected through the MCP server.

Every command#

This list comes straight from Azynote, so it always matches what azynote-ctl commands shows. Flags in square brackets are optional.

Driving the app#

Your workspace's data#

The same actions AI tools get through the MCP server, on the workspace open in the window.

Something else? Write to support@azynote.com.