Private AI meeting notes. No bot required.

Record any call without inviting a bot

YapStack captures your microphone and your computer's audio as two separate streams, so nothing joins the meeting and nobody gets a bot notification. Speech becomes text on your own machine as people talk.

• Capture modes: Mixed, Microphone Only, System Only.
• Lines are labelled You (mic) or Other (system audio). This is not speaker identification.
• English and 24 other European languages, detected automatically.

How bot-free recording worksMeeting transcription

Forgot to hit record? Rewind and keep it

While YapStack is capturing, it keeps the last few minutes of audio in memory. Start a session from that point in the past, and the audio you would have missed is saved and transcribed in the background while live transcription carries on. Buffered audio you never use is discarded and never written to disk.

Command Option R⌘⌥RCtrl Alt RCtrlAltRNew Session (Rewind)

• Shortcut ⌘⌥R on macOS, Ctrl+Alt+R on Windows.
• Keep the whole buffer or just the last few minutes.

How Rewind works

AI notes you can check against the recording

Ask the meeting a question, or use one click for a summary, key points, action items or minutes. Every answer carries timestamp citations that jump to that moment in the transcript and audio. When the AI renames, tags, files or saves something for you, you can undo it.

• One-click actions: Summarize, Key Points, Action Items, Meeting Minutes.
• Citations are timestamp chips. Click one to jump to that line and play it.
• Undo on every action the AI takes.
• Live Insights can comment on the transcript while the meeting is running. Off by default.

AI meeting notes with citationsLive Insights

Find anything you've ever said

One shortcut searches every folder, session, note, transcript line and dictation at once. Nested folders, tags and pins keep the archive in order. Give a folder some 'Context for AI' and the assistant knows what that folder is for.

Command K⌘KCtrl KCtrlKSearch everything

• Nested folders, tags, pins and drag-and-drop.
• AI suggests a folder for each session. Accept it or pick another.
• Copy or export a transcript as Markdown or plain text: all of it, just your side, or just theirs.

Search and organizeNotes and transcripts

Dictate into any app

Hold a shortcut, speak, let go, and the text is pasted at your cursor in email, Slack, a doc or your code editor. Speech recognition runs on your device, so dictation works offline. Give a slot an AI profile and it cleans up filler and false starts before pasting.

• Hold to talk, or switch to Toggle.
• Output to Paste, Clipboard Only or New Note. Paste goes through the clipboard and replaces what you had copied.
• Every dictation is kept in a searchable history.

Voice dictationAI dictation cleanup

Your AI, a local model, or none at all

Connect a cloud provider, or run a model on your own machine. Transcription never needs AI, so you can skip it entirely. If a connection fails, YapStack tells you. It never quietly sends your text somewhere else.

• OpenAI
• OpenRouter
• Ollama
• LM Studio
• llama.cpp
• vLLM
• Any OpenAI-compatible endpoint

• AI Connections and Profiles: pick a different model for chat, actions and each dictation slot.
• Use a local model and no text leaves your machine.
• Every AI feature is optional.

Bring your own AIRun a local LLM

Sync across devices through a server you run

Sync is off by default. Turn it on and your devices sync through a small relay you host yourself, with content encrypted on your device before it uploads. The relay stores ciphertext it cannot read.

• Relay never sees transcripts, notes, titles, audio, AI prompts or your password.
• New devices are approved by comparing a security code on both screens.
• The relay runs on your own server with Docker.

Encrypted syncSelf-host the relay