Can RecordCue post my meeting notes to Slack, Notion or Google Drive?

Not by itself, and that is deliberate. RecordCue records, transcribes on the device and files. A read-only command line prints where each finished recording and transcript live, and a script or local AI agent you run does the posting. The recipes on this page show the exact scripts for Slack, Google Drive, Obsidian, Apple Reminders and more.

Do I need to install a separate RecordCue CLI?

No. The command line ships inside the Mac App Store app at /Applications/RecordCue.app/Contents/MacOS/RecordCue --agent-cli. It has three read-only subcommands, latest, list --since and inspect, each with --json. It cannot start, stop or delete a recording, and it never uploads anything.

Does automation need an AI?

No. Filing a transcript into a shared Drive folder, copying notes into an Obsidian vault, a weekly digest of every call with your talk-time ratio, or a better transcript with Whisper are all AI-free. Summaries and action items need a model; the recipes use the local AI agent you already have, such as Claude Code or Codex, under your own account.

Does this work with ChatGPT in the browser or Claude.ai?

No. A browser chat cannot run commands or read files on your Mac. The recipes need a local agent session on the Mac, such as Claude Code, Codex CLI or a shell script. You can still attach a transcript to a web chat by hand.

RecordCue / Automate

Automate everything after the call.

RecordCue does the capture: the cue, both sides, a labeled on-device transcript, tidy filing. Then it gets out of the way. A read-only command line and plain files hand every finished recording to the tools you already run — and there is no AI service, account or upload inside the app to get in between.

1 · DetectA call starts. The cue appears.Zoom, Meet, Teams, Slack huddles, Discord, FaceTime, WhatsApp and about thirty more. One press records.
2 · CaptureBoth sides, or the screen.Mic and system audio in one file, or a window, area or whole display. macOS transcribes on the device, labeled by side.
3 · FilePlain files, plus JSON..m4a/.mp4, a .txt transcript, and metadata from the CLI or an optional .recordcue.json beside the recording.
4 · Your toolsAnything that runs a shell.Your AI agent, a launchd hook, Hazel, a script. Summaries, Slack, Drive, Obsidian, tasks — on your terms.

Three things make all of it possible.

None of them is a feature you switch on. They are how RecordCue is built, so every recipe below is a few lines of shell rather than an integration to wait for.

Plain files in plain folders

Every recording is an ordinary .m4a or .mp4 in ~/Movies/RecordCue, filed by service and named by date and time. The transcript is a .txt beside it, each passage marked Mic:, System: or Both:.

Nothing proprietary, nothing in a database. Any tool that reads files reads these — and keeps working if RecordCue is gone.

A signal that a recording is finished

The CLI lists only recordings that are safely written, so a job that asks it once a minute is a dependable hook. The transcript is attached to the same record when it is ready.

For Hazel and other watch folders, Settings → AI tools can also write a .recordcue.json companion beside each finished recording.

A read-only command line

latest, list --since 7d and inspect print JSON: paths, service, times, duration, capture modes, and a timeline of which side was speaking. Exit codes a script can trust.

It cannot start, stop or delete a recording, and it uploads nothing. That is what makes it safe to hand to any agent.

$ '/Applications/RecordCue.app/Contents/MacOS/RecordCue' --agent-cli latest --json
{
  "id":         "e1a8de54-…",
  "service":    "Google Meet",
  "startedAt":  "2026-09-08T09:07:01Z",
  "durationMs": 3257886,
  "capture":    { "microphone": true, "systemAudio": true, "video": false },
  "media":      { "path": "…/RecordCue/Google Meet/2026-09-08 18.07-19.01 Google Meet.m4a" },
  "transcript": { "path": "…/RecordCue/Google Meet/2026-09-08 18.07-19.01 Google Meet.txt" },
  "speakerActivity": { "format": "recordcue-rle-v1", "data": "rc1:100:.107M1.10M1…" }
}

Pick a recipe.

Each one is a complete, copyable setup, written so that you — or the agent you paste it into — can follow it end to end. They all start from the same hook and the same JSON, so they combine.

Trigger

Run anything when a recording finishes

The hook every other recipe builds on: a tested runner, installed with launchd, that runs your actions once per finished recording. Hazel and fswatch variants included.

Read the recipe →Share

A Slack summary when the call ends — no bot in the call

Decisions, action items and open questions, posted to a channel minutes after you hang up. Nothing joined the meeting.

Read the recipe →Tasks

Action items into Reminders or GitHub issues

Every "can you send me that" becomes a task with an owner, in the tracker you already use.

Read the recipe →Share

Meeting notes in a Google Drive folder your team can see

Notes and, if you choose, the recording itself land in a shared folder. Works the same with Dropbox, OneDrive and iCloud Drive.

Read the recipe →Notes

Every call becomes a note in Obsidian

Front matter, a summary, the full transcript, a link from the daily note. Local files into a local vault; nothing leaves the Mac.

Read the recipe →Transcribe

A better transcript with Whisper, still on the Mac

macOS transcription is instant and free. For the calls that deserve it, whisper.cpp re-transcribes the same file with a large model.

Read the recipe →Report

A weekly digest, and how much you talked

Every call of the week in one page, with your share of the talking computed from the speaker timeline. No AI required.

Read the recipe →Agent

Ask Claude Code or Codex about your last meeting

Install the Agent Skill once. Then "what did we decide on Tuesday" is a question, not a search through Finder.

Read the recipe →

Or let your agent read this instead of you.

Every recipe here is written to be executed, not just read. If you run Claude Code, Codex or another local agent on this Mac, paste this and answer its questions. It will read the pages, propose the files, and ask before writing anything.

Setup request for your agent

Works in any agent that can fetch a web page and run a shell on this Mac.

Read https://www.recordcue.app/automate/ and its recipe pages. Tell me which recipes fit how I work, ask me to pick one, then set it up on this Mac following that page exactly. Show me every file before you write it.

A browser-only chat cannot do this — it has no access to your files. The Agent Skill is the one-time install that lets a local agent find recordings on its own afterwards.

Written against the tools people already have.

The recipes name real tools and real paths, tested on a Mac, because “integrates with everything” usually means “integrates with nothing yet.” Swap any of them: the only RecordCue-specific line in every recipe is the one that asks the CLI where the files are.

Claude Code
Codex
Gemini CLI
Ollama
launchd
Hazel
fswatch
Shortcuts
Slack
Google Drive
Dropbox
iCloud Drive
Obsidian
Notion
Apple Reminders
GitHub
whisper.cpp
MacWhisper
jq
Python

What stays true, whatever you build.

RecordCue never records on its own.: Automation begins after a recording you started has finished. The cue is a prompt, not a trigger; ignoring it writes nothing.
Nothing leaves the Mac until your script sends it.: The app makes no network requests of its own. The moment a transcript reaches Slack, a model or a shared drive is a line you wrote, in a script you can read.
The CLI is read-only.: It cannot start, stop, delete, transcribe or summarise. A broken script can post the wrong thing; it cannot lose a recording.
Your files outlive the app.: Plain media, plain text, plain JSON in plain folders. Buy the capture tool once; keep the output forever.

Questions before you start.

Can RecordCue post my meeting notes to Slack, Notion or Google Drive?: Not by itself, and that is deliberate. RecordCue records, transcribes on the device and files. A read-only command line prints where each finished recording and transcript live, and a script or local AI agent you run does the posting. The recipes on this page show the exact scripts for Slack, Google Drive, Obsidian, Apple Reminders and more.
Do I need to install a separate RecordCue CLI?: No. The command line ships inside the Mac App Store app at /Applications/RecordCue.app/Contents/MacOS/RecordCue --agent-cli. It has three read-only subcommands, latest, list --since and inspect, each with --json. It cannot start, stop or delete a recording, and it never uploads anything.
Does automation need an AI?: No. Filing a transcript into a shared Drive folder, copying notes into an Obsidian vault, a weekly digest of every call with your talk-time ratio, or a better transcript with Whisper are all AI-free. Summaries and action items need a model; the recipes use the local AI agent you already have, such as Claude Code or Codex, under your own account.
Can a recording trigger a script the moment it finishes?: Yes. The CLI lists only finished recordings, so a launchd job that asks it once a minute is a reliable hook. RecordCue can also write a .recordcue.json companion beside each finished recording for Hazel and other watch-folder tools. The hooks recipe includes a tested runner that runs every action once per recording.
Does this work with ChatGPT in the browser or Claude.ai?: No. A browser chat cannot run commands or read files on your Mac. The recipes need a local agent session on the Mac, such as Claude Code, Codex CLI or a shell script. You can still attach a transcript to a web chat by hand.

The full JSON, exit codes and the speaker-activity format are in the agent CLI reference. Setup for the app itself is in the User Guide. RecordCue is free on the Mac App Store; one purchase removes the recording-length limit, and everything on this page is in the free version.