Fernscout

Imprint & privacy

Who runs this site, what it does with your data, and what it does not promise.

Last updated September 25, 2026

In short

  • Run by one person in Switzerland.
  • Everything lives on one server in Helsinki, Finland, with a second encrypted backup in Germany.
  • No tracking, no advertising, no third-party scripts. Cookies only to keep you signed in and remember your choices.
  • Outside services are used only for the feature that needs them, and only when somebody uses it.
  • The AI helper sees only what you choose to give it, and only after you said yes.
  • You can export or delete a journal at any time, and ask what is stored about you.

Who runs this

Fernscout™ is operated by Severin Lindenmann (sole proprietorship).

Severin Lindenmann, Feldweg 18, 5512 Wohlenschwil, Switzerland

Contact: agent@fernscout.ch

The controller for personal data under the Swiss Federal Act on Data Protection (FADP) and the GDPR is Severin Lindenmann, reachable at the address above.

Fernscout is run by one person. No particular availability is guaranteed; enquiries are usually answered within a few days.

Who is responsible for what. Severin Lindenmann is responsible for running this site: accounts and sign-in, the server and its logs, payments, and the services listed below. What goes into a journal — its words, photographs and the people on it — is decided by that journal's owner, who also decides who may read it. For that content this site works on the owner's behalf: it stores it, shows it to the people the owner allowed, and sends it where the owner asked.

The source code is public and can be read in full at github.com/severinlindenmann/fernscout. It is licensed under the Apache License 2.0: free to run, modify, self-host and redistribute, for as long as you like, with no restriction on competing with it. The Fernscout name and logo are not part of that licence — see the project's TRADEMARK.md.

What is stored, and why

The legal bases named here are the GDPR's, for readers in the EU. Swiss law asks for the same information without them.

WhatWhyBasisHow long
Journal content — text, photographs, dates, costs, the people on a tripIt is the journalContract with the ownerUntil the owner deletes it
Email addresses — of owners, of people on a trip, of invited readersAn address is the credential here: there are no passwords, sign-in sends a codeContractAs long as the account or invitation exists
Sessions and tokens — with the name of the browser that signed in, never its IP addressSo a browser stays signed in, and an agent can writeContractA reader's session a year, an agent's token seven days; revocable at any time
Readers and contacts — names, email addresses, phone numbers, postal addresses, and the owner's own notes on themSo an owner can send a day, a postcard or an invitation to the people they choseThe owner's legitimate interestUntil the owner removes them. Phone numbers and postal addresses are stored encrypted and are never shown to an agent
Reactions — which emoji was left on a day, under a random code your browser keepsSo a reader can react once and take it backLegitimate interestAs long as the journal
Conversations with the helper, and messages, photographs, documents and voice-note transcripts sent by WhatsApp or text messageSo an owner can carry on where they left offContractUntil the owner deletes them or the journal
Operator reading helper conversationsTo improve the helperLegitimate interest; the owner can object at any time (switch it off on their page)As long as the conversations exist
Location historySo a trip's map shows the roads really travelledConsent, per tripUntil the owner deletes it, by month or all at once
Page-open counts, for journals whose author switched them onSo an author can tell whether their readers readLegitimate interestAbout ninety days
Request log — the page asked for, the time, the browser's name; no IP addressTo run and defend the serverLegitimate interest14 days
Copies of mail this site sentTo see why a sign-in code did not arriveLegitimate interestTwo days, never backed up
iPhone app waitlist — an email address, and the language you asked forSo we can tell you once the app is on the App StoreConsentUntil the app is released, or you ask us to remove it
Unfinished signups — an email address you confirmed, the language of the page, and a phone number once you confirmed itSo you can continue where you stopped, on any device, without starting againSteps before a contract, at your requestEight days after you confirmed the email address, or until the journal is made — whichever comes first
Payments, credits and print ordersBookkeepingContract, and the law (Swiss bookkeeping rules)Ten years; see Deleting
Encrypted backups of all of the aboveSo a broken disk does not end a journalLegitimate interest14 days on the server, 7 days in the second copy

Only the owner of a journal talks to the helper — on the web because signing in requires it, and over WhatsApp because only a phone number the owner has already proven is ever bound to a journal. A message from any other number gets one fixed reply declining to help and a link to sign in, and never reaches the model. The operator reads helper conversations to see what to improve, and an owner can turn that off on their own page, at any time. Turning it off deletes nothing; it means nobody but them reads what is there or what comes next.

Location history is only there for a journal whose owner put it there: a file they imported (a Google or GPX export), or positions the iPhone app recorded during a trip they switched recording on for. It goes only to this journal's own server. Readers never see it: a trip's map shows a line drawn from it only for days they may already read, nothing from the last 24 hours, never the first or last 500 metres of a stretch, and nothing near places the owner marked private. The owner alone can be offered the name of the town a day was spent in, worked out on this server without asking anybody.

Only send what you have the right to share. A photograph with somebody else in it, or a document with somebody else's name on it, was never asked for their consent, and neither this page nor the software can give it on their behalf.

Giving an email address is needed to sign in. Everything else is optional, and each feature says what it needs when you use it.

On your device

There is no cookie banner, because nothing stored on your device is there to follow you: each item below is either needed for what you asked the site to do, or remembers a choice you made.

NameKindWhat it doesHow long
fs_sessionCookieKeeps you signed in to a journalA year
fs_identityCookieRemembers that you proved your email address, so you need not prove it again on every journalA year
fs.localeCookieThe language you pickedA year
fs.journalCookieWhich journal a welcome link was forA year
Theme, currency, speech language, dismissed noticesBrowser storageChoices you made on this deviceUntil you clear them
A reaction codeBrowser storageA random code, so the emoji you left can be taken backUntil you clear it
A push tokenBrowser storageSo this device can stop notifications it asked forUntil you clear it
Offline copiesBrowser cachePages you read, and whole trips you saved to read offlineUntil you clear them or remove the trip

There is no third-party analytics. No Google Analytics, no Plausible, no Matomo, no pixels, no advertising network, no third-party fonts or scripts. Nothing on these pages is loaded from anybody else's server.

The visitor count does not identify you. A journal's author may switch on a count for their own journal. It uses no cookie and no script. To tell two readers apart on the same day, the server makes a short code out of your internet address and your browser's name, mixed with a secret that is generated at random, kept only in memory, and thrown away every day. Your IP address itself is never written down, and the same person visiting tomorrow is counted as somebody new. The page you came from is deliberately not recorded, because it would show where a private link had been passed around.

Where the data goes

Everything is on one virtual server rented from Hetzner Online GmbH, in its data centre in Helsinki, Finland. Encrypted backups are kept on that server and in Hetzner Object Storage in Falkenstein, Germany. There is no CDN and no third-party database.

The services below are used only for what the row says, and only when somebody uses that feature. Nothing is passed to any of them for analysis, advertising or profiling.

ServiceWhereUsed whenWhat it receives
Proton AGSwitzerlandSign-in codes, invitations, notifications by emailThe recipient's address and the message
Twilio Inc.United StatesSomebody chose to get a code or a notification by text message, or texted the journalThe phone number and the text of the message
Meta Platforms Ireland (WhatsApp)Ireland, and the United StatesA reader asked to hear about new days by WhatsApp, or an owner's own proven number messages the journalThe phone number and the message, photograph, document or voice note
Apple (push notifications)United StatesA device running the iPhone app asked to be notifiedA device token, and the notification's title and link
Your browser's push service (Google, Mozilla or Apple)Depends on the browserA browser asked to be notifiedAn encrypted message it cannot read, and a device address
Anthropic PBCUnited StatesThe owner used the helper, after saying yes to it — see AI and voiceWhat the owner gave it for that one request
Deepgram Inc.Processed in the EU (US company)The owner spoke to the helper, or sent a voice noteThe recording, and its language
StripeIreland, and the United StatesSomebody bought creditsThe amount, the email address for the receipt, and the journal's name as a reference. Card or TWINT details are typed on Stripe's own page, never on this one
Stannp LtdUnited KingdomSomebody sent a printed postcardThe postcard's picture and message, and the recipient's name and postal address
Gelato ASANorwaySomebody ordered a printed photobookThe book, and the recipient's name, postal address and email address
Amazon Web Services (open elevation data)United StatesA photobook with a relief map was madeWhich map tiles are needed, which shows roughly where the trip went — nothing about a person
Komoot GmbH (Photon place search)GermanySomebody typed an address or a place to look up, or shared a locationThe words typed, or the coordinates. The request comes from this server, not your browser
Open-MeteoGermanyA journal asked what the weather was on a dayThe coordinates and date of that day. From this server
European Central BankGermanyA trip needed an exchange rateNothing: the request is for a whole published document

Transfers outside Switzerland and the EU. Finland, Germany, Ireland and Norway are covered by the EU's data protection rules, and the United Kingdom by an adequacy decision. For the services in the United States, the transfer rests on each provider's standard data processing terms, accepted when the account was set up rather than negotiated: for Meta and Anthropic the Swiss–US Data Privacy Framework; for Meta also its Business Data Processing Terms and Business Data Transfer Addendum; for Anthropic its commercial terms, which carry a data processing agreement. Deepgram, a US company, processes the recordings in the EU: this site sends them to its EU endpoint. Deepgram's data processing terms are not click-through in the same way, and whether one is in place for this instance is stated honestly: it is not, pending a request already sent to Deepgram's own privacy contact.

Meta plays two roles. For the message content you send, Meta acts on this journal's instructions and states that those messages are not used to target advertising. Separately, and on its own account, Meta keeps WhatsApp account data for platform safety and fraud detection; that is Meta's relationship with you as a WhatsApp user, described in WhatsApp's own privacy policy.

What Stripe does with payment data is covered by Stripe's privacy policy.

A link out is not a service on that list. Under a trip's map, and in some emails, there is "Open in Google Maps". It is an ordinary link carrying the coordinates of one place. Until somebody clicks it, nothing goes to Google. Click it and you are on Google's site, where Google's privacy policy applies and this operator has neither control nor sight.

AI and voice

The helper can write up a day, caption photographs, read a bank statement into a trip's costs, turn a photograph of travellers into drawn figures, and understand a spoken sentence. Each of those is asked for separately, in the app, before anything is sent: your words, your photographs, your voice, your bank statements. The question names the company it would go to. You can take a yes back at any time on the same page, and nothing is sent after that.

You said yes toSent toWhat is sent
Your wordsAnthropicWhat you typed or said, and the facts your own day already carries: its date, place, country, and how many photographs are on it and between which times
Your photographsAnthropicThe photographs you asked it to caption, or to turn into figures — which may show people
Your bank statementsAnthropicThe header row and five sample rows, so it can tell which column is which. Never the whole file
Your voiceDeepgramThe recording, and the language it is in

Nothing reaches either company unless the owner uses the helper. Reading a journal never does. A reader cannot cause a request, and nor can a stranger who messages the journal's WhatsApp number. No location history, no contacts and no addresses are ever sent to either.

The voice recording is never stored. It goes to the transcription service in one request and is dropped when the request ends; no copy is written to this server, to a backup or to an export. The text is kept, because the text is what was asked for. Both services are used only for the one request, and neither trains its models on it. Anthropic deletes what it was sent within 30 days, and keeps a request longer only when it is flagged for misuse, to check it. Deepgram is asked, with every recording, not to keep it or learn from it (its model-improvement opt-out); that is Deepgram's own promise, which this server cannot check.

The iPhone app

The app shows this same site, so everything above applies to it. What it adds:

  • Photographs. The app reads only the photographs you pick, and uploads them with their originals to your own journal.
  • Location, only for a trip you switched recording on for. iOS asks for "Always" access, and a message on the phone names the server and journal before recording starts. Positions go only to your own journal. Recording stops by itself the day after the trip ends, unless you chose to keep it running, and you can stop it at any time on the trip's page.
  • The microphone, only while you are recording something to tell the helper, and only after you allowed it. The recording goes to the transcription service as described in AI and voice and is never stored.
  • Notifications, only if you allow them, delivered through Apple.
  • Sharing into Fernscout from another app uses a sign-in kept in the phone's Keychain, valid for seven days.

Everything works if you say no. Without location, a trip's map has no recorded route; without the microphone, you type instead of speaking; without notifications, you get none. There is no tracking, no advertising identifier and no third-party analytics in the app.

Deleting

An owner can delete a trip, or a whole journal. Asking sends a confirmation email; nothing is removed until the link in it is followed. Then the content, the photographs, the readers and every row that belonged to the journal are deleted. A short note stays behind so the address can say the journal is gone rather than that it never existed. Copies in the backups expire within 14 days.

Payment records are kept. Swiss bookkeeping rules require them for ten years: what was bought, for how much, when, and the payment provider's reference. Names, email and postal addresses are removed from them when the journal is deleted. Stripe keeps its own record of the payment.

An owner can download everything as a zip of the journal's files and photographs from their own page, at any time.

A reader can ask the owner of a journal to remove them, or write to the address at the top of this page.

Your rights

You may ask what is stored about you, and have it corrected, deleted or handed to you in a format you can take elsewhere. You may object to anything done on the basis of legitimate interest, and withdraw a consent at any time without affecting what was done before. No decision about you is made automatically. Write to the address at the top of this page; you will get an answer within thirty days.

You can also complain to the Swiss Federal Data Protection and Information Commissioner (edoeb.admin.ch) or, in the EU, to the data protection authority where you live.

Weather and exchange rates

The weather is looked up by this server, not by your browser. It is asked once, when a day is written, and stored in that day's own file, so reading a page that shows the weather sends nothing to anybody.

Weather data by Open-Meteo.com, used under the Creative Commons Attribution 4.0 licence. The changes made: temperatures are shown in whole degrees and rainfall to one decimal place, and Open-Meteo's weather code is drawn as one of seven pictures rather than printed. Open-Meteo is itself a front end onto the national weather services — MeteoSwiss, the DWD, ECMWF, NOAA, Météo-France, the JMA and others, each under its own open licence, listed on their licence page. A reading a traveller recorded themselves names whoever took it and is credited to nobody else.

Exchange rates from the European Central Bank, whose reference rates may be reproduced with the source acknowledged. This server fetches the whole published table; the ECB is not told which trip, currency or date is of interest. The change made: the ECB quotes every currency against the euro, so a rate here has been cross-divided into the journal's own currency and rounded to six figures. A costs page that used a looked-up rate names the ECB and the date beneath the totals. A rate somebody typed in themselves carries no citation.

What is not promised

Fernscout is offered without any guarantee of a particular availability.

Keep your own copy. Journals are backed up regularly, but a backup can fail and a restore can be incomplete. If a trip matters to you, download it from your own page.

Security. Personal data is protected with appropriate technical and organisational measures — credentials are hashed, tokens expire, private trips are refused rather than merely hidden, and the code is reviewed for it. If a data breach happens anyway, the competent authority and the people affected are informed as the law requires.

Links. The operators of other sites linked from here are responsible for their content.

Liability. Liability is excluded to the extent the law permits. This does not apply to damage caused intentionally or by gross negligence, or to mandatory claims under data protection and product liability law.