Data & trust
Who owns what, where it lives, what the AI sees, and how we keep it all GDPR-compliant. The answers your data protection officer will ask for, before they ask.
Ownership
It’s yours
Content, members and research results belong to you. Leave any time with a full export.
Hosting
Frankfurt
Application and database run in Frankfurt, on infrastructure inside the EU.
Isolation
Your own instance
Every partner gets its own deployment, database and cache. No shared tables with other partners.
GDPR
Private by default
No tracking without consent, no email tracking, a data processing agreement with every contract.
Who owns what
Everything you put into your agora, and everything it collects for you, is yours: library resources, events, city and country profiles, member accounts and profiles, newsletters, chat logs and your brand. We process it only to run your platform, on your instructions.
What the research agents find is public information about organisations, projects, events and knowledge, with the source linked. We claim no rights to it.
The software stays ours, and you get every improvement as it ships. Your partner contract gives you the right to use it for as long as you stay.
When you leave, you get a full export: your database as an SQL dump, your content as CSV and JSON, and your uploaded files. We then delete our copies within 30 days, and you can keep your domain.
Where your agora lives
| What | Service | Where |
|---|---|---|
| Application | Vercel | Frankfurt (fra1) |
| Database (members, resources, settings) | Neon Postgres, on AWS | Frankfurt (eu-central-1) |
| Uploaded images and files | Vercel Blob | Delivered through a global CDN |
| Transactional email and newsletters | Amazon SES | Frankfurt |
| Rate limiting and cache (one per partner) | Upstash Redis | Frankfurt |
| Error monitoring (no IPs, emails or cookies) | Sentry | EU region (Germany) |
| Visitor statistics (no cookies) | Umami, run by us | Frankfurt |
| Tools you choose to connect (optional) | Notion, Airtable or Google Sheets | Where that tool stores your workspace |
| Video streaming (if you publish videos) | Mux | USA (EU–US Data Privacy Framework) |
| AI models (assistant, research, summaries) | Google Gemini; Anthropic and OpenAI for some tasks | Provider data centres, not pinned to the EU |
Several of these companies are based in the USA, even where your data sits in Frankfurt. Each one is bound by a data processing agreement, and transfers rest on the EU–US Data Privacy Framework or the EU’s standard contractual clauses. You get the full sub-processor list with the contract, and we tell you before we add a new one.
Your own instance
Each partner platform is its own deployment with its own database, its own cache, its own connections to outside tools and its own sign-in cookies. A sign-in on one platform is rejected by every other, and an outage or a mistake on one platform stays on that platform.
Some infrastructure is shared to keep it affordable: the file store (each partner in its own folder), error monitoring, and our accounts with the AI and email providers. City and country indicators from Eurostat, the OECD and the World Bank are pooled across platforms too; they contain no personal data.
What the AI sees
Research agents
They receive your topic, your territories and the searches they plan, and they send search queries to search and AI services (Google Gemini, Exa, Serper, Perplexity). They read public web pages about organisations, projects, events and knowledge. They don’t look for or profile people, and they never see your member data.
The assistant
A member’s question and the matching excerpts from your library go to Google Gemini to write the answer, with sources. We use Google’s paid API, so Google doesn’t use these questions to train its models. Conversations are stored in your database, linked to the member’s account or to a daily-changing code instead of the visitor’s IP address, so you can fix wrong answers.
Voice
Speech streams straight from the browser to Google to generate the spoken answer. Neither the audio nor a transcript is stored.
Everything else
Summaries, tagging and member matching run on the same paid APIs (mostly Gemini; OpenAI and Anthropic for some tasks). These providers don’t train on paid API data, and neither do we: your content never becomes anyone’s training set. AI requests are processed in the providers’ own data centres, which aren’t pinned to the EU, so don’t put sensitive personal data in the assistant.
How we keep it GDPR-compliant
Clear roles
For your portal you are the controller and we are your processor. We sign a data processing agreement (Auftragsverarbeitungsvertrag, Art. 28 GDPR) with the partner contract, with the sub-processor list and our technical and organisational measures attached.
Private by default
Visitor statistics run on our own cookieless Umami. Anything that tracks, such as Google Analytics if you want it, loads only after the visitor consents. No tracking pixels or tracked links in emails. Double opt-in for newsletters. Where we need to tell visitors apart, IP addresses are replaced by keyed hashes that change every day. Member profiles are public only if the member switches them on.
Members’ rights, built in
Members can download their own data and delete their own account. Your editors can find, correct and export member data in the admin, so you can answer access and deletion requests in minutes instead of days.
A privacy policy that fits
We hand you a privacy policy and cookie table written for your exact setup, based on the one live on housingcoop.eu, for your counsel to review. When we change how the platform handles data, we tell you what to update.
Security
TLS everywhere, data encrypted at rest by our providers, passwords stored only as hashes, optional two-factor sign-in, sign-in tokens encrypted, role-based admin access, and request limits against abuse. We run on infrastructure providers with SOC 2 and ISO 27001 certifications. We’re a small team and don’t hold certifications ourselves, so we tell you exactly what we do instead.
If something goes wrong
If a data breach affects your platform, we inform you without undue delay, with what we know, so you can meet your 72-hour notification duty.
Need it in writing?
Ask for our data processing agreement, the full sub-processor list or a call with your data protection officer. We’d rather answer the hard questions before you sign.
This page describes the platforms we run for partners. For this website itself, see our privacy policy and imprint.