Privacy policy

Last updated: 17 November 2026 · Version 1.20

This translation is provided for information only. Only the German version is legally binding.

This policy explains which personal data we process when you use loomira.ai and the Loomira portal, for which purposes, on which legal basis and for how long. It applies to website visitors, prospects and business customers.

1. Controller

The controller for the processing described here is Staqmind UG (haftungsbeschränkt), Gilgaustraße 36, 51149 Köln, Germany, represented by Mirko Rossbach, Montassar Zaroui.

A data protection officer has not been appointed; the requirements of § 38 BDSG are not met. Please direct privacy requests to datenschutz@loomira.ai.

Where the Loomira assistant answers calls on behalf of our customers, the respective customer is the controller. We process that data as a processor under Art. 28 GDPR on the basis of a data processing agreement.

2. What we collect — and why

Server logs

When you visit our website we process technically necessary access data (IP address, timestamp, requested page, user agent) to keep the service running and to fend off attacks. Legal basis: Art. 6 (1)(f) GDPR (legitimate interest in secure operation). Deleted after 30 days at the latest.

Web analytics (Plausible)

To understand how our website is used we employ Plausible Analytics, a privacy-first analytics service by Plausible Insights OÜ (Västriku tn 2, 50403 Tartu, Estonia; an EU company — data is processed and stored exclusively on servers in the EU). Plausible works without cookies and does not track visitors across websites or devices: the IP address and user agent are used only transiently to derive a daily-changing pseudonymous identifier and are not stored. We only ever see aggregated statistics (pages visited, referrers including campaign parameters in links, countries, device types) and counts of individual steps such as requested demo calls, started and completed sign-ups and started orders, without reference to any person. Legal basis: Art. 6 (1)(f) GDPR (legitimate interest in measuring reach); a data processing agreement under Art. 28 GDPR is in place. You may object to this processing at any time (Art. 21 GDPR).

YouTube videos

Some pages embed videos from our YouTube channel. The provider is Google Ireland Limited, Gordon House, Barrow Street, Dublin 4, Ireland. We use a two-click solution in YouTube’s privacy-enhanced mode (youtube-nocookie.com): until you click the play button, we show only a preview image served from our own server, and your browser does not connect to YouTube or Google. Only your click loads the YouTube player. Google then receives in particular your IP address, browser and device information, the address of the page you are on and information about how you use the video; YouTube may also store or read information on your device (for example cookies or local storage). If you are signed in to a Google account, Google can link the visit to that account. We store nothing ourselves; Google is responsible for any further processing, including in the USA. Google LLC is certified under the EU-U.S. Data Privacy Framework (adequacy decision, Art. 45 GDPR). The legal basis is your consent, given by clicking the play button (Art. 6 (1)(a) GDPR, § 25 (1) TDDDG). It covers only that video; you can withdraw it at any time with effect for the future by reloading the page (the preview image returns) and deleting the youtube-nocookie.com data in your browser. Without your click, no data is sent to YouTube. For details, see Google’s privacy policy.

Demo call

If you request a demo call via our website, we use your phone number solely to place that call and to prevent abuse (rate limiting via hashed values). Technical logs can contain your phone number for up to 30 days. Legal basis: Art. 6 (1)(b) GDPR. No marketing use takes place.

Demo call: protection against abuse

When you request a demo call, we also check on our servers, using an offline country database, which country your IP address belongs to, so that the demo only calls numbers in the country you are browsing from. Your IP address is not sent to the database provider or to any other third party for this. We refuse requests made through anonymizing services (VPN, proxy, Tor). Legal basis: Art. 6 (1)(f) GDPR (legitimate interest in preventing toll fraud).

IP geolocation by DB-IP , licensed under CC BY 4.0 .

Booking a demo

If you book a demo on our website, we process your first and last name, your work email address, your company and your phone number and, if you provide them, your notes and any further answers you give in the booking form, to arrange and hold the video call; we use your phone number only to reach you with questions about the appointment. The booking is created by our scheduling provider Cal.com, Inc. (USA), which sends the calendar invitation and reminders on our behalf. Your browser does not connect to Cal.com: our server transmits the booking. Cal.com is bound as a processor under Art. 28 GDPR; the transfer to the USA is based on EU standard contractual clauses (Art. 46 GDPR). Legal basis: Art. 6 (1)(b) GDPR. Deleted after 12 months at the latest, unless a contract follows.

Contacting us

If you contact us by email, we process your details to handle the enquiry. Legal basis: Art. 6 (1)(b) GDPR for (pre-)contractual enquiries, otherwise (f). Deleted after 12 months at the latest, unless statutory retention duties apply.

Account and contract data

For your Loomira account we process registration and contract data (name, email address, company, configuration). Login runs through our self-hosted Zitadel instance in the EU. Legal basis: Art. 6 (1)(b) GDPR. Deleted 30 days after the contract ends, unless statutory retention duties apply.

Payment data

Payments are processed by Stripe. Payment data (e.g. card details) is processed exclusively by Stripe and is not stored by us. Legal basis: Art. 6 (1)(b) GDPR. We retain invoices for ten years pursuant to § 147 AO.

Transactional email

System emails (e.g. account confirmation, invoices) are sent via Scaleway (Paris region, EU). Legal basis: Art. 6 (1)(b) GDPR.

AI phone calls (on behalf of our customers)

Where the Loomira assistant answers calls for a business, we process call content on behalf of that business (Art. 28 GDPR). Call audio is not stored — it is processed transiently for real-time handling only. Transcripts are anonymised after 30 days; we store the caller’s phone number so the business can call back (shown shortened in the call list) and anonymise it together with the transcript. For analytics and for recognising returning callers we use only a salted hash. At the start of each call the assistant identifies itself as an AI. The respective customer, as controller, is responsible for the lawfulness of this processing; its legal basis arises from the customer’s relationship with the callers (usually Art. 6 (1)(b) GDPR).

We provide our customers with a template privacy notice for callers: View the template

3. Cookies and local storage

We use strictly necessary cookies only (§ 25 (2) no. 2 TDDDG). They are essential for operating the website and the portal and require no consent:

  • i18n_redirectedRemembers your language choice (German, English, French, Italian or Spanish).
  • loomira-sessionKeeps you signed in to the portal (encrypted session, only after login).
  • loomira-auth-flowSecures the login and signup flow.
  • loomira-csrfProtects forms against cross-site request forgery.

A cookie consent banner is therefore not required: our web analytics (Plausible, section 2) works without cookies and without cross-site tracking, we use no advertising services, we self-host our fonts, and third-party content (YouTube videos) loads only once you click its play button yourself (section 2).

The portal additionally stores interface preferences (e.g. dark mode) locally in your browser; this data never leaves your device.

4. Processors and recipients

We use carefully selected service providers that process personal data on our behalf, contractually bound under Art. 28 GDPR. The application and its databases run in the AWS Frankfurt region (eu-central-1), where the data is stored. Requests to the language model and the knowledge-base search are handled by our providers in further EU data-centre regions named in the list of sub-processors. Some providers also process data outside the EU: speech synthesis (ElevenLabs) currently runs on a global endpoint, so processing in the US is possible, and transfers to the US are possible for payments (Stripe) and telephony (Telnyx); which providers these are, where they process data and on what basis is set out in the list of sub-processors. Transfers to third countries take place only under the safeguards of Chapter V GDPR (an adequacy decision, such as the EU-US Data Privacy Framework, or EU standard contractual clauses under Art. 46 GDPR); for providers with a US parent company, EU standard contractual clauses and supplementary assessments (transfer impact assessment) are in place.

  • Amazon Web Services EMEA SARL (Hosting) — Cloud hosting: compute (ECS Fargate, EC2), databases (Aurora PostgreSQL), cache, storage, encryption (AWS KMS), logging and monitoring (Region eu-central-1 (Frankfurt am Main, Germany))
  • Amazon Web Services EMEA SARL (Amazon Bedrock) — AI inference via Amazon Bedrock: embeddings and reranking for the knowledge base (Cohere Embed v4, Cohere Rerank 3.5) and a backup language model (Anthropic Claude) used if Google’s language model is unavailable. Processes knowledge-base content (website text, uploaded documents, your own texts), search queries from calls and chats and, for the backup language model, the conversation history. Amazon Bedrock also runs the language model (Anthropic Claude) of the AI setup assistant, which customers use to set up and change their assistant in the portal by chat. The setup assistant processes what the customer writes in the chat, the details of their business, their assistant’s settings and content read from their website during setup; it does not receive call transcripts. AWS runs the models itself: the model providers receive no data, and Amazon Bedrock neither stores inputs and outputs nor uses them for training (EU: eu-central-1 (Frankfurt am Main, Germany), eu-north-1 (Stockholm, Sweden), eu-south-1 (Milan, Italy), eu-south-2 (Spain), eu-west-1 (Ireland), eu-west-3 (Paris, France). The EU inference profile spreads requests across these regions; reranking runs in eu-central-1 only)
  • Google Cloud EMEA Limited — AI inference via Vertex AI: language model (Gemini) for call and message replies and for sorting and summarising calls, and embeddings for the knowledge base (previous method and fallback). During a website import, the language model (Gemini) reads the text of the customer’s website and suggests details for the setup from it (such as trade, services and opening hours). For the AI setup assistant in the portal, Gemini is the backup model if Anthropic Claude on Amazon Bedrock is unavailable; it then processes the same data (chat messages, business details, the assistant’s settings, content read from the website) but no call transcripts (Vertex AI: EU multi-region (language model), region europe-west1, Belgium (embeddings) and region europe-west3, Frankfurt am Main, Germany (backup model of the AI setup assistant))
  • Soniox, Inc. — Speech recognition (speech-to-text) for the AI phone assistant (EU endpoint (processing in the EU))
  • ElevenLabs, Inc. — Speech synthesis (text-to-speech) — processes only the spoken response text (Global API endpoint — processing may take place outside the EU (USA); a switch to the EU residency endpoint is planned)
  • Telnyx Ireland Limited — Telephony: phone numbers, call routing (SIP) and sending SMS (if enabled) (Calls via Frankfurt (EU), EU API endpoints; transfers to the US possible (Telnyx’s main processing location, e.g. for call records, SMS and support))
  • Stripe Payments Europe, Ltd. — Payment processing and invoicing (Dublin, Ireland (transfers to the US possible))
  • Scaleway SAS — Transactional email (e.g. account confirmation, invoices) and email notifications to the customer’s team, such as the transcript of a call once it ends (Paris, France (fr-par region))

We operate authentication (Zitadel) and the voice infrastructure (LiveKit) ourselves in our own AWS environment in the eu-central-1 region (Frankfurt) — they are not separate sub-processors.

Our web analytics provider Plausible (section 2) processes website usage data only and is not part of the sub-processor chain for customer call data.

Our scheduling provider Cal.com (section 2) processes demo bookings only and is not part of the sub-processor chain for customer call data.

Services a customer connects to Loomira itself are operated by the customer or under its own contract with the respective provider — for example calendar and booking systems, spreadsheets, CRM systems and address books, ticketing and help-desk systems, industry software (e.g. for property management or trades), team chat, messaging channels such as WhatsApp, automation services and the customer’s own webhooks and MCP servers. Loomira exchanges data with these services only on the customer’s instruction, given by setting up the connection; they are not sub-processors of Loomira. Where such a service processes data depends on its provider and may be outside the EU.

For the WhatsApp channel, a customer connects its own WhatsApp Business account and contracts directly with Meta for the use of the WhatsApp Business Platform. Meta processes the messages on the customer’s behalf under the customer’s own contract with Meta; Meta is not a sub-processor of Loomira. Processing outside the EU is possible in this context.

We maintain the complete, current list — with purpose, location and third-country basis — publicly: List of sub-processors

5. Calendar and spreadsheet integrations (Google Calendar, Google Sheets, Microsoft 365 / Outlook)

If you connect your Google Calendar, Google Sheets or your Microsoft 365 / Outlook calendar in the portal, you grant Loomira access on Google’s or Microsoft’s own consent screen (OAuth 2.0); we never see your password. The assistant then checks free times in the calendar you select and books appointments into it during calls. If you allow callers to change their bookings in the booking tool, the assistant can also move or cancel appointments that Loomira itself created; it never changes or deletes any other event.

  • What we read: the list of your calendars (names and identifiers), so that you can choose the calendar to book into, and the free/busy times of the selected calendar(s). The identifier of your primary Google calendar is your Google account’s email address; we use it to show which account is connected. The free/busy data itself is used only during the call to offer available times and is not stored beyond the call; the times offered appear in the call transcript like the rest of the conversation. To move one of its own appointments, the assistant retrieves the events that overlap that appointment and uses only their start and end times, to work out which time becomes free; all other details, such as titles, are discarded immediately and are never stored, logged or passed on. For a Microsoft account we also read your basic profile (name, email address and account identifier) to show which account is connected.
  • What we write: one event per booked appointment in the selected calendar, containing the caller’s name, the caller’s phone number (where it was transmitted or the caller gave it) and the notes from the booking. We add no attendees and send no invitations.
  • Google Sheets call log: if you connect Google Sheets, Loomira creates one spreadsheet in your Google Drive and adds one row per call: time, assistant, call status and duration, the caller’s phone number (where it was transmitted), and the name, phone number, email address, request and notes the assistant recorded. Loomira can only open files it created itself (Google’s drive.file permission) and cannot see any other file in your Drive. To show which account is connected, we read your Google account’s email address.
  • Calendar and spreadsheet data is used solely to provide these features (booking appointments, keeping the call log). It is never sold, never used for advertising, never used to determine creditworthiness or for lending, never used to train AI models and never passed on to third parties for their own purposes.
  • Loomira staff do not view your calendar or spreadsheet data, including the times offered as they appear in transcripts, unless (a) you have given your explicit, documented consent to view specific data, for example in a support request, (b) it is necessary for security purposes, such as investigating abuse or a bug, (c) it is required by law, or (d) the data has been aggregated and anonymised for internal operations.
  • What we store: the connected account’s email address, the identifiers of the calendars you selected, the identifier of the spreadsheet Loomira created and, for each appointment Loomira booked, the event’s identifier, date and time and the caller’s name, so that the booking can be found again. The caller’s name is deleted once the appointment lies further back than your retention period for call data.
  • The access tokens issued by Google or Microsoft are stored encrypted (AWS KMS) in the AWS region eu-central-1 (Frankfurt am Main) and deleted when you disconnect the integration.
  • How we protect it: Google user data is encrypted in transit (TLS) and at rest (AWS KMS), and only authorised staff can access our systems, with role-based permissions and multi-factor authentication (see section 9). The language model never receives the titles or contents of your events, only the free times, and every booking change is checked on our servers and limited to appointments Loomira created, so instructions a caller speaks cannot make the assistant read or change any other event.
  • Deletion: when you disconnect, Loomira immediately deletes the access tokens and the connection details, including the account’s email address, and revokes its access at Google, unless another Loomira workspace still uses the same Google account. The calendar identifiers stored in a booking tool are deleted with that tool; the booking records are deleted with your account data after the contract ends (section 6). You can ask us to delete your data at any time using the contact details in section 1.

Who receives Google user data

We share, transfer or disclose data received from Google APIs, raw or derived (for example the free times calculated from your free/busy data), only to the following recipients and only as far as needed for the features described above:

  • Amazon Web Services, our hosting provider (region eu-central-1, Frankfurt am Main): stores the connection and appointment data listed above and, if Google’s language model is unavailable, runs our backup language model (Amazon Bedrock, EU regions), which then receives the free times instead.
  • Google Cloud, our language-model provider (Vertex AI, EU multi-region): during a call, the language model receives the free times so that the assistant can offer them to the caller.
  • ElevenLabs, our speech-synthesis provider: turns the assistant’s reply, which can include the times offered, into speech; processing may take place in the USA (EU standard contractual clauses).
  • Telnyx, our telephony provider: carries the call and, where enabled, the text message confirming the appointment.
  • Scaleway, our email provider: sends emails to your team, such as call summaries or a request to reconnect that names the connected account.
  • Google itself: appointments are written to your own Google Calendar and call-log rows to the spreadsheet Loomira created in your Google Drive, on your instruction.
  • The caller: hears only the free times offered and the details of their own appointment, never the content of your other events.
  • Members of your Loomira workspace: see which account is connected, the appointments Loomira booked and the call transcripts.
  • Tools and channels you set up yourself: the transcript and summary of a call, which can mention the times offered and the appointment booked, go only where you send or retrieve them yourself, for example via post-call delivery to Slack, Microsoft Teams, your CRM or your own webhook, or through our developer API or MCP server. Your calendar list, your other events and your Google account details are never sent there. You can switch these off at any time.
  • Authorities or courts, only where we are legally obliged to disclose data.
  • A buyer or successor, only in a merger, acquisition or sale of assets of our company, and only with your explicit prior consent.

The service providers named above (Amazon Web Services, Google Cloud, ElevenLabs, Telnyx and Scaleway) process the data as our processors under Art. 28 GDPR, only on our instructions and only to provide their service to us; they may not use it for their own purposes, including training AI models (see section 4 and the list of sub-processors). Beyond this, we do not share, transfer or disclose Google user data to anyone: we never sell it and never pass it to advertising platforms, data brokers or information resellers.

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

Google publishes this policy at developers.google.com/terms/api-services-user-data-policy .

We do not use, retain or transfer raw or derived user data received from Google Workspace APIs (Google Calendar, Google Sheets and Drive) to develop, improve or train generalized or non-personalized AI or machine-learning models, and we do not train any AI model with it at all. Our language-model and speech-synthesis providers process it only to produce the assistant’s reply during the call and may not use it to train their models. The use of raw or derived user data received from Workspace APIs will adhere to the Google User Data Policy, including the Limited Use requirements.

You can revoke Loomira’s access at any time in the portal (Integrations → Disconnect) or directly with your provider: Google at myaccount.google.com/permissions , Microsoft personal accounts at account.live.com/consent/Manage , Microsoft work or school accounts at myapps.microsoft.com .

Google and Microsoft are your own providers: they process your calendar and spreadsheet data under your contract with them and act on your instruction, given through your own account. They are not sub-processors of Loomira (see section 4); where they process the data, including outside the EU, is governed by your agreement with them.

Legal basis: Art. 6 (1)(b) GDPR (performance of the contract with you) for the connection and the tokens. The booking details and call-log entries of callers we process on your behalf as a processor (Art. 28 GDPR) and transfer them to your calendar or spreadsheet provider on your instruction.

6. Retention periods

Unless stated otherwise above, the following periods apply:

  • Server logs30 days at the latest
  • Contact enquiries12 months
  • Account and contract data30 days after the contract ends
  • Invoices10 years (§ 147 AO)
  • Business correspondence6 years (§ 257 HGB)
  • Call audionot stored
  • Call transcriptsanonymised after 30 days
  • Caller data on bookingsdeleted 30 days after the appointment
  • Deleted assistantsrestorable for 30 days, then deleted for good together with the contents of their calls
  • Technical logs that can contain phone numbersat most 30 days
  • Deleted data in backupsup to 14 days

7. Your rights

You have the rights under Art. 15 to 21 GDPR:

  • Access to your stored data (Art. 15 GDPR)
  • Rectification of inaccurate data (Art. 16 GDPR)
  • Erasure (Art. 17 GDPR)
  • Restriction of processing (Art. 18 GDPR)
  • Data portability (Art. 20 GDPR)
  • Objection to processing based on legitimate interests (Art. 21 GDPR)

You may withdraw any consent you have given at any time with effect for the future.

To exercise these rights, contact datenschutz@loomira.ai.

You also have the right to lodge a complaint with a data protection supervisory authority. Competent for us: State Commissioner for Data Protection and Freedom of Information of North Rhine-Westphalia, Kavalleriestr. 2–4, 40213 Düsseldorf, www.ldi.nrw.de.

8. Automated decision-making

No profiling within the meaning of Art. 4 no. 4 GDPR takes place.

Where the assistant automatically accepts or declines appointment requests on behalf of a customer, this is done to perform or initiate a contract between the caller and the customer (Art. 22 (2)(a) GDPR). The safeguards of Art. 22 (3) GDPR apply: callers can request human handling at any time and contest the decision — the customer remains reachable through the usual channels.

9. Data security

We take technical and organisational measures pursuant to Art. 32 GDPR, including:

  • Encrypted transport (TLS) for all connections
  • Hosting in EU data centres (AWS region eu-central-1, Frankfurt am Main); deviations by individual sub-processors are listed in the sub-processor list
  • Encryption of stored data (AWS KMS)
  • Role-based access on a need-to-know basis with multi-factor authentication
  • Salted hashes instead of plain-text phone numbers for recognising returning callers
  • Data minimisation: only the response text reaches speech synthesis; call audio is not stored
  • No storage of payment data — card details are processed exclusively by Stripe (PCI-DSS)

10. Data breaches

In the event of a personal data breach we notify the competent supervisory authority within 72 hours (Art. 33 GDPR) and inform affected persons without undue delay where there is a high risk (Art. 34 GDPR). Where we act as a processor, we notify the responsible customer without undue delay (Art. 33 (2) GDPR).

11. Changes to this policy

We update this policy when the legal situation or our processing changes. The version published here applies; you can find its date at the top of this page.