# Samurai Scribe

> Markdown-first AI meeting notetaker for macOS and Windows. Transcription never leaves your computer. Meetings are saved as Obsidian-flavoured Markdown you own, with a built-in AI assistant, skills, connections, and an MCP server for Claude, Codex, Cursor and Gemini CLI.

Samurai Scribe is a desktop application that records meetings without a bot, transcribes them and separates speakers on your own CPU or GPU, and writes every meeting into a folder of plain Markdown and audio (a "vault"). A built-in AI assistant answers questions across all your meetings and notes, citing the exact moment it draws on. Transcription and speaker detection always run on your computer. Summaries, dictation polish and the assistant run on a local model by default; a hosted model (Anthropic, OpenAI, Gemini, OpenRouter or a custom address) is optional, and the first time you pick one the app says what will be sent and to whom. No account is required. Current version: 0.65.0 (November 2026).

## Key Capabilities

- **No meeting bots**: Captures system audio and microphone at the OS layer. Nothing joins your Zoom, Google Meet or Teams call.
- **Local speaker detection**: On-device speaker separation, including several people sharing one microphone in a room, with speaker renaming and analytics (talk time, word counts, duration). A Speakers card beside the transcript shows who spoke and for how long; drag one speaker onto another to merge them, and names from a meeting's calendar invite are suggested.
- **Voice prints**: Confirm a speaker's name once and the app keeps a voice print for that person — stored in your own vault, never in an account — and recognises them in later meetings, both after processing and live while you record. Only the people a meeting expects (you, its invitees and anyone you added to it) are named automatically, marked unconfirmed; anyone else who sounds like someone you know is offered as a suggestion, and otherwise the speaker keeps a placeholder name (Mic Speaker or Sys Speaker 2 in a recording, Speaker 2 in an imported file) until you name them. Naming one line by hand names the other lines in the meeting that sound like that person. Each person is recognised by up to three voice clips, each removable. On a call, a track can be locked to one or several people. You can clear a person's voice print and keep the person, or delete them entirely. Free.
- **People notes**: Each person you name gets a note in your vault's People folder, so `[[Name]]` in a transcript resolves in Obsidian and every person has backlinks to the meetings they spoke in. A Speakers page lists everyone you have named. Turn the notes off or choose the folder in Settings → Vault.
- **Mic vs. system channel isolation**: Records microphone and system audio as separate aligned tracks with echo cancellation.
- **Live subtitles**: The transcript is written as the meeting happens, and kept: running the meeting afterwards goes on from it to the summary rather than transcribing again.
- **The assistant**: Ask about your meetings and notes in a panel beside any meeting (Cmd+J / Ctrl+J) or full width in the Projects view. Answers cite the exact moment; clicking a citation opens the meeting at that moment, with the playhead there. It suggests notes and edits that wait for review by default, with four modes (Ask, Review, Accept edits, Auto), undo and chat rewind. It remembers facts about you (About You) and per-project memory, accepts meetings, notes and images with @, and drafts emails for you to open in your mail app; it never sends anything itself. Free.
- **Live meeting help (Pro)**: While recording with live subtitles on, four one-click actions: Catch me up, Help me answer (from your notes and past meetings), Ask a good question, Add something useful. Optional running notes. It answers when asked; it does not interrupt.
- **Companion window**: Ctrl+Option+N (Win+Alt+N on Windows), pressed in any app, turns Samurai Scribe into a small window over your other apps for the meeting being recorded, or the one on screen, with every tab of the meeting and the assistant. While recording, one button puts the current time into your Notes as a link to that moment. Hidden from screen sharing by default; the key can be changed in Settings → Hotkeys.
- **Skills**: Plain-English instructions for repeated tasks, as SKILL.md files in the open Agent Skills format. Run by hand on a meeting or calendar meeting, with / in a chat, or as a step of a profile's summarization (free). Automations run a skill after meetings, before calendar meetings or on a schedule while the app is open (Pro). Skills can take arguments (for example a language). Ships with four built-in skills (Meeting summary, Name speakers, Polish transcript, Translate transcript, which translates a finished transcript into any language you name, for review); a public, MIT-licensed skills library lives at https://github.com/samuraiscribe/skills (meeting prep brief, weekly review, my action items, one-on-one tracker, decision log, sales call notes, follow-up email, interview scorecard, coach me, lecture study guide, action items to Linear).
- **Connections (beta)**: The assistant and skills can work in Linear, Notion, Jira & Confluence, Todoist, monday.com, ClickUp and Attio through each service's own remote MCP server. You sign in on the service's page; each action is set to off, ask first, allowed, or allowed in automatic runs (Pro). What a tool is given goes to that service under its terms.
- **AI Apps (MCP server)**: Claude Desktop, Claude Code, Codex, Cursor and Gemini CLI can search and read your meetings and notes through a helper that Settings → AI Apps → Connect an App sets up, while Samurai Scribe is running. Its tools search, list and read meetings, transcripts, documents, notes, notebooks and skills, speaker stats and upcoming meetings, and suggest creates, edits, moves, deletes, transcript corrections and renamed people, which wait for review in the app by default. Folder rules (Can edit / Read only / Hidden) decide what each app sees. The app writes an AGENTS.md file to the vault describing its layout. The connected AI app sends what it reads to its own AI service.
- **Vaults (files you own)**: One folder per meeting holding a meeting note, transcript, documents and recordings as Obsidian-flavoured Markdown and audio. Transcripts have a heading per line (time and speaker) so Obsidian can link to a moment. External edits are picked up, even while the app was closed. The files are the record: the app rebuilds everything else from them (Settings → Vault → Rescan Vault). Notebooks group meetings and notes on one subject.
- **Projects and Files**: A Projects view with full-width chats grouped into projects, and a Files browser for the whole vault with an Obsidian-style live-preview editor (tables, task lists, callouts, links, images, properties).
- **Attachments**: PDFs, Word, PowerPoint, text files and images can be attached to a meeting before processing; images are read by a local model by default. The assistant also reads and searches PDFs, Word files and decks kept anywhere in the vault (decks in slide order, with their speaker notes) and takes them with @.
- **The Dictionary**: Your own spellings of names, products and jargon, applied wherever speech becomes text (meetings, live transcripts, dictation). Lists are Markdown notes in the vault's Dictionary folder, with their own page in the app; add a word from Replace All in a transcript, a selected word's menu, or a dictation in your history. Once its model is downloaded it also listens for your terms in the audio, in English and 17 other languages. One list's first ten entries are free; more entries and more lists need Pro.
- **Compare models**: Settings → AI Models → Transcription → Compare models runs two transcription models side by side on a sample, one of your meetings or a passage you read aloud, and shows which was faster and which got fewer words wrong. On Free, a Pro model can be downloaded and tried there.
- **System-wide dictation**: Hold fn (macOS) or Ctrl+Win (Windows) and speak; text is typed into whichever app has focus. Speech is transcribed on-device, and dictated text is kept out of clipboard history. The Clean up polish prompt is free; the other polish prompts and custom prompts are Pro.
- **Scriptable summaries (Rhai)**: Summaries can be driven by user-editable Rhai scripts that read the transcript, attached notes and files, and the linked calendar event, and write any number of documents.
- **Importing recordings**: Drop an existing recording onto a meeting, or pick it from your files, and it is transcribed, speaker-labeled and summarized like a live one (free). Audio: WAV, MP3, M4A, FLAC, OGG, AAC, WMA, Opus, AIFF. Video (its audio track): MP4, MOV, MKV, AVI, M4V, WMV, FLV, 3GP, WebM. Import Bulk Recordings scans a folder or many files and imports each as its own meeting: one recording at a time is free, several at once is Pro.
- **Online media import**: Import and transcribe audio or video from a link (YouTube, TikTok, Instagram, Vimeo and 1,000+ other sites).
- **Calendar**: Read-only Google Calendar (OAuth) or ICS sync; calendar meetings can record themselves (Pro). When Zoom, Teams, Meet or another call app starts using the microphone, the app offers to record; for calls not on the calendar it can instead record automatically (Pro).

## Privacy and network

- Transcription and speaker detection always run on the computer. Audio leaves it only in a recording the user chooses to contribute (Settings → Support & Troubleshooting → Contribute recordings), which asks them to confirm that everyone in it agreed.
- With local models (the default), recordings, transcripts, summaries and notes are never transmitted anywhere, unless the user sends a meeting with a bug report or contributes it.
- Optional, user-chosen: hosted AI models, connections, and AI apps connected over MCP. Each sends only to the provider or service the user picked.
- Logs and the troubleshooting file leave out meeting titles, people's names, calendar events, account names and file and folder names.
- A daily update check carries five anonymous fields (app version, OS, tier, install week, activation level). Usage data is on by default and can be switched off; no per-install identifier is ever sent.
- Full list of every request the app can make: https://samuraiscribe.com/network

## Models

- **Transcription**: Whisper and NVIDIA Parakeet, on your computer, using the GPU where it can (Metal on a Mac, Vulkan on Windows), free on every tier. Settings → AI Models → Transcription → Advanced → Run transcription on: Automatic, GPU or CPU only.
- **Speaker detection**: on your computer.
- **AI**: a built-in model the app downloads (Qwen 3.5 4B, about 3 GB), or Ollama, LM Studio, or a hosted model, chosen in Settings → AI Models.
- **Scripting**: Rhai.

## Pricing

- **Free ($0, no time limit, no account)**: Unlimited transcription with Whisper Tiny, Base and Small and NVIDIA Parakeet v3 and v2, on the GPU or the CPU; the Dictionary (your spellings of names and jargon, one list of ten entries); speaker detection and voice prints that recognise people across meetings; AI summaries; the assistant; skills run by hand and the skills library; connections (beta); AI Apps (MCP); dictation with Clean up polish; custom scripts; vaults; calendar integration and analytics.
- **Pro ($19 one-time until Monday, November 16 at 11:59 PM PST (the beta price, with a discount code at checkout), $39 list, plus sales tax or VAT where the buyer's country charges it; lifetime updates, 2 machines)**: Everything in Free, plus live meeting help; skills that run by themselves (automations); Whisper Medium, Large V3 Turbo and Large V3; Parakeet Ultra (NVIDIA Parakeet retrained by Moondream); the full set of dictation polish prompts and custom prompts; automatic recording of calendar meetings and of calls not on your calendar; a processing queue, so several meetings run one after another and the queue survives a restart; Import Bulk Recordings of more than one recording at a time (one at a time is free), which turns a folder, a selection of files or links into meetings dated from the recording; and every Dictionary entry past the first ten, in as many lists as you like. 7-day money-back guarantee.
- **Trial**: None. Since 0.65.0 new installs start on the free version. On Free, a Pro model can be downloaded and tried side by side with yours in Settings → AI Models → Transcription → Compare models.
- **Enterprise (contact us)**: Pro for everyone on a team, plus priority email support, an onboarding call, and help with security reviews (we fill in the questionnaire and walk IT through the network disclosure). Priced per team by email: support@samuraiscribe.com. Coming soon, not yet in the app: admin controls (lock installs to local models; turn off cloud models, connections and AI apps), a shared team vocabulary, and shared skills and templates. Details: https://samuraiscribe.com/enterprise

## Platform Compatibility

- **macOS**: Apple Silicon (M1 or later), macOS 13+ (DMG). No Intel build yet.
- **Windows**: Windows 10/11 64-bit (code-signed EXE installer; adds the WebView2 runtime if missing). GPU acceleration via Vulkan with CPU fallback.
- **Linux**: Not yet.

## Docs

How to use the app, one page per topic. Each link is the page as Markdown; the same page without `.md` is the web version. All of them in one file: https://samuraiscribe.com/llms-full.txt

### Start here

- [How Samurai Scribe works](https://samuraiscribe.com/docs/how-it-works.md): The path a meeting takes in Samurai Scribe, from recording or import to transcript, speakers, summary, your vault and the assistant.
- [Your first meeting](https://samuraiscribe.com/docs/first-meeting.md): From first launch to a transcript, named speakers, a summary and a first question to the assistant. The setup steps, the downloads, and what to press.
- [Transcription](https://samuraiscribe.com/docs/transcription.md): Choose the language and model Samurai Scribe transcribes with, try two models side by side, and pick whether it runs on your graphics card or processor.
- [The Dictionary](https://samuraiscribe.com/docs/dictionary.md): Your own spellings of names, products and jargon, used wherever speech becomes text in Samurai Scribe, in meeting transcripts, live captions and dictation.
- [Glossary](https://samuraiscribe.com/docs/glossary.md): What Samurai Scribe means by vault, meeting, run, profile, step, script, skill, automation, project, notebook, Dictionary, AI app and its other words.

### The assistant

- [The assistant](https://samuraiscribe.com/docs/assistant.md): Ask Samurai Scribe about your meetings and notes. Where to open the assistant, what it reads, citations, images, saving answers, memory and which model answers.
- [Changes, modes and undo](https://samuraiscribe.com/docs/assistant/changes.md): What the assistant, skills and AI apps may change without asking. The four modes, reviewing a change, a meeting's changes, undo, rewind and deleted notes.
- [Live meeting help](https://samuraiscribe.com/docs/assistant/live-help.md): With Pro, ask the assistant about the meeting you're in while it's recorded. The four quick actions, your own questions, running notes and the companion window.

### Projects and files

- [Projects and chats](https://samuraiscribe.com/docs/projects.md): How the assistant's chats are kept. Projects and Home, starting, moving and finding chats, a project's memory, and what reaches your other computers.
- [Files and the editor](https://samuraiscribe.com/docs/files.md): The Files pane shows your vault as a tree beside the assistant. Open files in tabs, make, move and delete notes and folders, and write in the note editor.
- [Folder access](https://samuraiscribe.com/docs/folder-access.md): What the assistant and connected AI apps may do in each folder of your vault, Can edit, Read only or Hidden. The defaults, how rules apply, and what each guarantees.

### Your vault

- [The vault](https://samuraiscribe.com/docs/vault.md): Where Samurai Scribe keeps your meetings, a folder of Markdown and audio you own. What is in it, what you can edit, what to leave alone, and how to sync it.
- [Using Samurai Scribe with Obsidian](https://samuraiscribe.com/docs/vault/obsidian.md): Keep your meetings in your Obsidian vault. Which folder choice to make, what meetings look like in Obsidian, links, tags, Dataview queries, plugins and syncing.

### Skills and automations

- [Skills](https://samuraiscribe.com/docs/skills.md): Plain-English instructions for tasks you repeat. The built-in skills, translating a transcript, running a skill, the shared library and what a skill may change.
- [Writing a skill](https://samuraiscribe.com/docs/skills/writing.md): Make your own skill in the app's editor or as a SKILL.md file. The fields, the settings the app reads, arguments, what a skill can do, and how to make it work well.
- [Automations](https://samuraiscribe.com/docs/automations.md): Have a skill run by itself after each meeting is processed, before calendar meetings, or on a schedule. Automations need Pro and run while Samurai Scribe is open.

### Profiles and scripts

- [Processing profiles](https://samuraiscribe.com/docs/profiles.md): A profile is the recipe a run follows, from transcribing and telling voices apart to the steps that write the meeting up. Choosing, changing and making them.
- [Scripting](https://samuraiscribe.com/docs/scripting.md): Scripts decide exactly how a profile writes up a meeting, and can update notes across your vault. Where they live, the editor, the step's settings, and a first script.
- [Scripting reference](https://samuraiscribe.com/docs/scripting/reference.md): Everything a Samurai Scribe script can use in scripting API 1.6: the meeting, transcript, notes, model, your vault, calendar, standard library, limits and versions.
- [Scripting cookbook](https://samuraiscribe.com/docs/scripting/cookbook.md): Complete Samurai Scribe scripts to copy and change, each checked with Validate in 0.65.0: languages, long meetings, notes in your vault, stats and more.

## Links

- Website: https://samuraiscribe.com
- Docs: https://samuraiscribe.com/docs/ (every page also as Markdown; all of them in one file at https://samuraiscribe.com/llms-full.txt)
- Release notes: https://samuraiscribe.com/releases/
- Skills library: https://github.com/samuraiscribe/skills
- Network disclosure: https://samuraiscribe.com/network
- Enterprise: https://samuraiscribe.com/enterprise