Legal
Privacy Policy
Clowhi is built to be local-first: Clowhi never uploads your audio, and the website collects only what it needs to sell and deliver the app. This policy explains what we do and do not process, and the rights you have under the General Data Protection Regulation (GDPR).
1. Who is responsible (controller)
The data controller for personal data processed through this website is:
Reconnect Academy BVErembodegemstraat 31/15, 9300 Aalst, Belgium
Company number: 0738.879.286 · VAT: BE 0738.879.286
Privacy contact: support@guerrillabuild.com
2. The Clowhi app is local-first - and you are in control of it
Reconnect Academy BV is the controller for the data described in section 3: your order, your payment record and your support emails. For everything the app handles on your Mac - recordings, transcripts, meeting notes, dictation history - we are neither the controller nor a processor. We never receive that data and have no way to access it. If you use Clowhi privately, that processing is normally outside the GDPR's scope as a purely personal activity. If you use it for work, your organisation is the controller for those recordings and transcripts, including towards the other people in the meeting, and is responsible for having a lawful basis, informing participants and answering their requests. Section 2 below is given so you can meet those duties, not because we process that data.
The Clowhi application runs on your own Mac. It is designed so that your data stays with you:
- Clowhi never uploads audio. Recording and WhisperKit transcription run on your Mac. Clowhi captures microphone and system audio, not screen video. Your own vault or watch-folder sync can transport files you choose to sync; Clowhi does not operate that transport.
- Text is sent only when you switch on an AI feature, and then only to the provider you configured: Anthropic's Claude API under your own account, or Ollama on this Mac. The finished Markdown note - summary, transcript and any personal meeting notes - is written to the folder you choose. Nothing is sent to us.
- Audio goes to the Trash, not a shredder. After a recording is processed, the source audio is moved to your Trash so you can still recover it. A failed summary does not change this once a transcript exists. Emptying the Trash is up to you, and your own backups or sync may still hold copies.
- Failed recordings are kept, not deleted. If transcription fails permanently
and no transcript was produced, the audio is preserved in a
failed/folder so you do not lose it. Interrupted or retryable recordings stay in the app's local recovery folder until they are processed.- The finished Markdown note contains the summary or pending-summary state, transcript, and any personal meeting notes. It is written to your chosen folder.
- Sessions, recovery files and local note-history references can remain on the Mac.
- After normal processing, source audio moves immediately to the recoverable Trash; delayed audio retention is not implemented.
- A summary failure alone does not preserve audio once a transcript exists.
- Permanently failed watch-folder imports with no transcript are preserved under the source folder's
failed/folder; temporary failures leave the audio in place for retry. - Interrupted recordings and retryable processing failures keep captured audio in the local session/recovery directory; a terminal empty/silent recording with no transcript is preserved under its session's
failed/folder. - Failed writes or cleanup can retain recovery files.
- Trash is not secure erasure, and backups or sync may retain copies.
- Dictation history is on by default. When you use dictation, the app can keep up to 50 active dictation text entries for up to 7 days. This history is local and is never sent to any AI provider, including local ones, or to any other endpoint. You can disable it in Settings or clear it in the Dictation pane; cleanup moves entries to Trash, so those limits do not erase copies in Trash or backups. The app reports storage or cleanup failures.
- My Day (Beta) is a spoken work log on your Mac. When you log an update, Clowhi records your microphone, transcribes the recording locally and keeps the text, your corrections, follow-ups and day recaps as local files in Clowhi's Application Support folder. Once an update's text is saved, its audio moves to the recoverable Trash; if transcription or saving fails, the audio stays on your Mac so you can retry. Capture, manual organizing, follow-ups and the local day recap send nothing to any server. The optional setting “Use Claude for My Day” is off by default and separate from meeting summaries and dictation. If you turn it on after reading its disclosure, the text of new work updates and the same-day context you allow is sent from your Mac to the AI provider you configured: with Claude under your own account and API key, with Ollama only to Ollama on this Mac. Earlier updates, follow-ups from other days and saved meeting actions are sent only when you select them and approve a preview of exactly what will be sent. Turning the setting off stops queued sends; anything already sent cannot be taken back. Your updates often mention other people, which is their personal data, and turning the setting on is not their consent. My Day never reads or uses your dictation history.
- Your Vocabulary list stays on your Mac. Your words, the ways Clowhi may mishear them, and your snippets are stored only on this Mac, in Clowhi's Application Support folder. Clowhi uses the list to correct transcripts and, where your speech model supports it, to help that local model recognise your words. The list itself is never sent to your AI provider or any other endpoint. Your words reach your AI provider only as part of text that this policy already describes being sent, such as the corrected transcript when you use summarization (sent together with the summary format, your custom instructions and, by default, your personal meeting notes, as described below) or a corrected work update when AI for My Day is on, never as the list itself.
- Model and tokenizer downloads. Clowhi downloads required files from Hugging Face and reuses complete local copies. It downloads only the model files on its own curated list, at fixed versions, and checks every file before it uses it. The host receives normal connection metadata, such as IP address and requested files, but no meeting content. Runtime preparation may need a tokenizer download even when model weights are installed.
- Automatic update checks. The app contacts clowhi.com through the Sparkle updater for signed-update metadata, and downloads an update when you accept one. Those requests carry no recording, transcript or personal notes, and no account identifier - we have no accounts. They do generate the ordinary server logs described in section 3. Sparkle asks you once whether it may check automatically. No accounts. No product-usage analytics and no advertising trackers. The app checks our server for updates, and that request reaches our server logs like any web request.
- Claude summaries use your own API key. When summarization is configured, Clowhi sends the transcript text, selected summary format and custom instructions to Anthropic under your own API account. Personal meeting notes are also sent when “Include my notes in the summary” is on, which is the default. Turn it off to keep those personal notes local while still writing them into the Markdown note. That switch does not stop the transcript or summary instructions from being sent. Audio and dictation history are never included. The API key is stored in macOS Keychain. Without a key, local transcription still works and a later summary retry can send the transcript after a key is configured. That request goes directly from your Mac to Anthropic, under your own account and your own API key: we are not a party to it, we hold no data from it, and we have no processor agreement with the provider covering your content. Where and how the provider processes or transfers your data is set out in the provider's own terms and privacy documents, which you should read. If you use Clowhi for work, your organisation is the controller for what it sends, and is responsible for checking what a transcript may contain - meeting transcripts routinely include other people's personal data and can include sensitive categories.
- Ollama on this Mac instead of Claude. Clowhi supports two providers: Anthropic's Claude API, under your own API key, and Ollama running on this Mac, which you install and run yourself. If you choose Ollama, Clowhi sends the same text only to Ollama at 127.0.0.1 on this Mac, never to another computer, and lists only models that Ollama runs locally; Ollama needs no API key. If you use summarization with Ollama selected, the same request (transcript text, summary format, custom instructions and, when that setting is on, your personal meeting notes) goes only to Ollama on this Mac. Ollama is software you install and run yourself, under its own license; Clowhi needs no account or API key for it, and there are no provider usage charges. With Ollama as your provider the My Day setting reads “Use Ollama for My Day” and its text goes only to Ollama on this Mac. Nothing is sent to us either way.
- Privacy information and permissions. The app stores the version of the privacy information shown or acknowledged locally. That is a UX record, not blanket legal consent or consent from meeting participants. Microphone and system-audio permissions enable meetings; Accessibility enables optional dictation insertion, Input Monitoring applies to the Fn/Globe shortcut, and Notifications are optional.
Clowhi does not send meeting content to our website. Optional AI requests and your own vault sync are separate from the website, purchase and support flows described below.
3. What the website processes
Depending on how you interact with us, we may process:
- Order and billing data: name, email address, billing address, country, and VAT/company details where relevant.
- Transaction records: order, invoice, payment status, refund and delivery records.
- Support data: the content of messages you send us and technical information you choose to include.
- Server and security logs: limited technical data such as IP address, timestamp and requested URL, generated by our hosting provider.
Your full payment card details are collected and processed directly by Stripe, our payment processor, not by this website. We receive only the transaction result and limited billing information needed to fulfil and account for your order.
4. Purposes and legal bases
We process the data above to:
- process your payment and deliver the download - performance of a contract;
- provide support and handle refunds - performance of a contract and our legitimate interest in helping customers;
- prevent fraud and keep the service secure - legitimate interest;
- meet tax, invoicing and accounting duties - legal obligation;
- establish, exercise or defend legal claims - legitimate interest.
We do not send marketing email by default. Any optional marketing would rely on your separate consent, which you could withdraw at any time.
5. Who we share data with (processors and recipients)
We share personal data only with parties that help us run the service, or where the law requires it:
- Stripe - payment processing. See stripe.com/privacy.
- Vercel - website hosting and delivery. Vercel generates the server and security logs described above. See vercel.com/legal/privacy-policy.
- Apple - our support and privacy mailbox runs on iCloud Mail, so Apple processes the email you send us. See apple.com/legal/privacy.
- Professional accounting or legal advisers, and public authorities, where required by law.
We do not sell your personal data. As of this version the website does not use any analytics, advertising or third-party tracking tools.
6. Cookies and analytics
This website sets no cookies at all and embeds no third-party tracking. Fonts are served from our own domain. If we ever add analytics or marketing tools, we will update this policy and put the required consent mechanism in place before those tools run.
7. International transfers
Some of our providers (for example Stripe and Vercel) may process data outside the European Economic Area, including in the United States. Where that happens, the transfer is covered by an appropriate safeguard such as the EU Standard Contractual Clauses or an adequacy mechanism, as described in that provider's own documentation.
The Hugging Face model and tokenizer downloads that the app makes from your own Mac are described in section 2 above.
8. How long we keep data
We keep order and invoice records for the periods required by Belgian tax and accounting law (currently up to 7 years). Support and security data are kept only as long as reasonably necessary for the purpose they were collected for, or for the duration of a possible legal claim, after which they are deleted or anonymised.
9. Your rights
Under the GDPR you have the right to:
- access the personal data we hold about you;
- have inaccurate data corrected;
- have data erased where the law allows;
- restrict or object to certain processing;
- receive your data in a portable format;
- withdraw any consent you have given, without affecting past processing.
To exercise any of these rights, contact support@guerrillabuild.com. You also have the right to lodge a complaint with the Belgian supervisory authority:
Gegevensbeschermingsautoriteit / Autorité de protection des donnéesDrukpersstraat 35 / Rue de la Presse 35, 1000 Brussels, Belgium
www.gegevensbeschermingsautoriteit.be
10. Changes to this policy
We may update this policy as our service, providers or legal obligations change. The version in force is the one published here, carrying its "last updated" date.
11. Contact
Privacy questions or requests, and general support: support@guerrillabuild.com.