Legal

Privacy Policy

Last Updated: November 6, 2026

The "Local-First" Guarantee

Samurai Scribe was built on the principle that your private meetings should stay private. Your audio stays on your device unless you choose to contribute a recording, and transcription and speaker detection always run on your own computer's hardware. With local AI models (the default), your transcripts, summaries and notes stay there too. Meeting text goes anywhere else — to a cloud AI model, a connected service or an AI app — only if you set that up. Every network request the app is capable of making is listed in our network disclosure.

1. Overview

This Privacy Policy describes how Samurai Scribe ("we", "us", or "our") collects, uses, and handles your information when you use our website and desktop application.

2. Data Handling (The App)

The Samurai Scribe desktop application is designed to be a local-first tool.

  • Audio Data: We do not have access to your microphone input or audio files. They are processed entirely in memory or on your local disk, unless you choose to contribute a recording (see 2.2).
  • Meetings and Transcripts: Your recordings, transcripts, summaries and notes are stored as files in a vault folder you choose on your own computer. We do not have a "cloud sync" feature, and we cannot see the content of your meetings unless you send us one with a bug report (see 2.1) or contribute it (see 2.2). If the vault folder you choose is synced by another service, such as iCloud Drive, OneDrive or Dropbox, your files go wherever that service takes them.
  • AI Models: By default, transcription, speaker detection, summaries, dictation polish and the assistant run on models downloaded to and run on your computer, and no data is sent to external AI providers. If you choose a hosted model (for example Anthropic, OpenAI, Google Gemini or OpenRouter) for summaries, polish or the assistant, the text that task works on is sent to that provider, using your own API key and under that provider's terms. Your audio is never sent.
  • Connections: If you connect a third-party service in the app (such as Linear or Notion), what one of its tools is given when it is used is sent to that service under its terms. Nothing is sent to a service until you connect it.
  • AI Apps: If you connect an AI app (such as Claude Desktop, Claude Code, Codex or Cursor), it reads your meetings from Samurai Scribe on your computer and sends what it reads to its own AI service, under its own terms. Samurai Scribe itself sends nothing in that process.
  • Anonymous Usage Data: Once a day the app checks for updates. With anonymous usage data on (the default; you can turn it off in Settings → Updates), that request also carries your app version, operating system, whether you are on Free, a Pro trial or Pro, the week you installed, and whether you have transcribed or summarized anything yet. It carries no identifier and nothing from your meetings.
  • Downloads: The app downloads the files it runs on: AI models, from our model host, which passes the request on to GitHub or Hugging Face (Whisper models come from Hugging Face directly); app updates, from GitHub; and, only if you import meetings from links or install it in Settings → Online Media, the video downloader yt-dlp and its JavaScript runtime Deno, from their official releases on GitHub. With "Keep the downloader up to date" on, an import also asks GitHub whether a newer yt-dlp exists. These requests carry nothing from your meetings. Our network disclosure lists every one.

2.1 User-Initiated Support Requests

If you report a bug from the application, you fill in a form: what happened, and an email address for the reply. You may also include recent logs, with meeting titles, people's names, email addresses and the names of your files taken out; a description of your computer's setup (its hardware, installed models, permissions, the app's settings, and what kind of place each folder is, never its path) and, if the problem is about a particular meeting, that meeting's text, chosen by name. All of these are optional and can be inspected before anything is sent.

This is never automatic. Nothing is sent until you press "Send report". A report is filed as an issue in Samurai Scribe's private GitHub repository and kept with that support request, so the problem can be traced if it comes back. What you wrote and your email address are stored in that issue; your email address is used only to reply to you. The troubleshooting file waits in Cloudflare's storage until the report is looked at, then moves onto the issue; one that is never looked at is deleted after 30 days. A meeting's text is read to find the problem and then deleted; it is never kept with the support request. Please don't include patient or client details in a report. To have a report and its files deleted, email support@samuraiscribe.com.

Licensed users can also send a testimonial the same way. It is quoted on samuraiscribe.com only if you tick the box allowing it.

2.2 Contributed Recordings

You can choose to send us meetings to help test and improve transcription and speaker detection (Settings → Support → Contribute recordings). This is never automatic. Nothing is sent until you confirm that everyone who speaks in them agreed, and press "Send". For each meeting you pick, the app sends its audio and its transcript, with each named speaker replaced by Person 1, Person 2 and so on. Words said in the meeting stay as they are, names spoken aloud included. The meeting's title, notes and summaries are never sent. You may also include recent logs and a description of your computer, as for a bug report; these are read when the recordings are added to our test set, then deleted.

Contributed recordings are kept in storage only the developer can read, to test future versions of the app, and are deleted if you ask at support@samuraiscribe.com. Your email address is used only to reply to you. Please don't contribute meetings with patient details. Exactly what is sent is listed in our network disclosure.

2.3 Google Calendar

If you connect a Google account in Settings → Calendar, the app asks Google for read-only access to your calendars (Google's calendar.readonly permission) and for your account's email address. It cannot change, add or delete anything in your calendars, and it cannot read your mail, contacts, files or anything else in your Google account.

  • What the App Reads: Every five minutes while you are connected, the app reads the list of your calendars, to find your main calendar and the others you show in Google Calendar, and the events in them from six hours ago to seven days ahead. From each event it keeps the title, the start and end times, whether it is confirmed, your own reply, the organizer, each guest's name, email address and reply (meeting rooms are left out), the location, the description, the video call link, and the event's ID and link in Google Calendar. Anything else Google sends with an event is not kept. From your sign-in it keeps your account's email address and Google's ID for the account.
  • How It Is Used: To show your upcoming meetings; to offer to record a meeting when it starts (or, with Pro and if you turn it on, to start recording by itself); to name a recording after its meeting and link the two; to suggest who is speaking from the guest list and spell their names right; to give a meeting's summary its agenda and guest list; and to let the assistant and your skills answer questions about your meetings and help you prepare for them. Your email address names the account in Settings and tells you apart from the other guests.
  • Where It Is Kept: Only on your computer. The events are kept in the app's own database and replaced at every sync. For the meetings it offered to record, or that you chose a setting for, it keeps the title and start time for up to 30 days, so it never asks twice. Google's sign-in token, which lets the app go on reading your calendars, is encrypted with a key tied to your computer and kept in your system's keychain, with an encrypted copy in the app's own data folder; the short-lived tokens used for each request are kept only in memory. Your vault records which Google account it is connected to (the email address and Google's ID for the account, in its .samurai folder), never the token. When a recording is linked to an event — because you chose the event, or because you started it while that was the only meeting under way — a copy of the event's details is saved with the meeting in your vault, so the meeting still shows who was invited after the event is over. If you name a speaker after a guest on the invite, that guest's email address is saved with the person in your vault. If your vault folder is synced by a service such as iCloud Drive, these files go wherever that service takes them.
  • Who It Is Shared With: Nothing from your Google account is sent to Samurai Scribe's servers, and we never sell it, use it for advertising, or use it to train AI models. It leaves your computer only in these cases, each of which you set up:
    • A hosted AI model. With a hosted model chosen for summaries or the assistant (see "AI Models" above), the calendar details a task works on go to that provider with the rest of the task, using your own API key and under that provider's terms. A meeting's summary is given its event's title, its description and its guests' names (an email address where the invite gives no name). The assistant and your skills can read your events for the next two days and the event linked to any meeting: the title, time, organizer, guests' names and email addresses, location and description. A skill run before a meeting is given that meeting's details, and with "Keep running notes during meetings" on, the agenda in an event's description is used for the notes. With local models, the default, none of this leaves your computer.
    • AI apps. An AI app you connect (see "AI Apps" above) can read the same calendar details the assistant can, and sends what it reads to its own AI service, under its own terms.
    • Connected services. If the assistant or a skill uses a tool of a service you connected (see "Connections" above), what it puts into that tool, such as a guest's name, goes to that service.
    • A bug report. If you include a meeting in a bug report (see 2.1), its text can include the calendar details its summary was given.
  • How to Remove It: In Settings → Calendar, click the delete button beside the account ("Remove this calendar") and confirm. The app asks Google to revoke its access, deletes the sign-in token and the cached events, and removes the account from your vault. If your computer is offline at that moment, Google is not told, so remove the access from your Google Account as well. Meetings already linked to the account's events keep their copy of the event, as the app says when you remove it; deleting a meeting moves it, with that copy, to your computer's Trash. You can remove Samurai Scribe's access at any time at myaccount.google.com/permissions; the app then can no longer read your calendars, and the events it already has stay on your computer until you remove the account in the app too. None of this data is on our servers, so there is nothing for us to delete.

Samurai Scribe's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.

3. Information We Collect (The Website)

When you visit our website or purchase a license, we collect minimal information:

  • Order Information: When you purchase a Pro license, our payment processor (Lemon Squeezy) collects your email address and billing information. This is required to issue your license key and handle tax compliance.
  • License Validation: If you activate a Pro license, the app periodically contacts our licensing provider (Lemon Squeezy) to verify that your license key is valid. Activating includes your license key and a hashed hardware identifier; later checks include your license key and the activation id our licensing provider gave your computer. Both are used to prevent piracy. It does not include any information about your transcripts or audio, and the free tier makes no licensing requests at all.
  • Website Analytics: This website uses Cloudflare Web Analytics to count page views and see which sites link to us. It sets no cookies, does not fingerprint your browser, and cannot follow you to other sites. We see totals — how many people visited a page, which site referred them, and which country they were in — never individuals. This applies to this website only: the app's own requests, none of which carry your audio, are set out in our Network Disclosure.
  • Download Counts: Clicking a download button passes through our own server so we can count downloads and see which link brought you here. It records the date, your operating system, the referring campaign tag, and the country the request came from. If you arrived through a tagged link, your first page also sends one landing record per browser tab with the same details. Each also records whether the request came from a page on this site and whether it looks automated (a headless browser, a declared bot or link previewer, or a cloud server), so crawlers and email scanners can be counted apart from people; the second is stored as a single word, never your browser details or network. No cookie is set and nothing identifies you. The tag is kept in your browser tab's session storage until the tab closes, and nowhere else.
  • Server Logs: Our own servers — for downloads, update checks, model downloads, the skills list, bug reports and contributed recordings — run on Cloudflare, which keeps a log of each request to them, with the network details Cloudflare attaches to any request, which can include your IP address. We keep these logs for up to seven days, only to find out why something broke.

4. Third-Party Services

We use Lemon Squeezy as our Merchant of Record. They handle payments, taxes, and license key delivery. You can view their privacy policy on their website. We do not store your credit card information on our servers.

Services you choose to use from the app — Google, if you connect a Google Calendar (see 2.3), hosted AI model providers, connected services such as Linear or Notion, and AI apps such as Claude Desktop — are separate from us and handle what you send them under their own privacy policies. We do not receive that data.

5. Your Rights

We do not store your recordings or transcripts, so for your use of the app there is nothing for us to "delete" or "export", with two exceptions: a bug report you send us, and anything you included with it (see 2.1), and recordings you contribute (see 2.2). We delete either on request. For purchase-related data, you may contact us at any time to request the deletion of your email from our licensing system.

6. Contact Us

If you have questions about this policy, please contact us at:
support@samuraiscribe.com