Trust
Where your data lives
Database, servers and queues are in the EU. The language model runs at Anthropic and OpenAI in the US. Below, per component: where it runs, what the language models retain and how long we keep things.
Updated October 25, 2026
Hosting
Where it runs
Per component: which provider delivers it and where the servers are.
App and website
Vercel · Frankfurt (fra1)
The app and this site run at Vercel. The server functions are in Frankfurt; images and scripts come from a global network (CDN). Vercel is a US company.
Database and login
Supabase · Frankfurt (eu-central-1)
Accounts, workspaces and Clyde's memory live in a database at Supabase in Frankfurt. Supabase manages that database; we do not host it ourselves.
Automations and scheduling
Hetzner · Nuremberg
Our own server at Hetzner, a German company, runs n8n for the automations, the knowledge base and Postiz for scheduling posts.
Queues
Upstash · Frankfurt (eu-central-1)
Tasks that run later and temporary data are kept at Upstash in Frankfurt. Upstash is a US company.
Language model
The language model runs in the US
Clyde writes mostly with Claude from Anthropic, and uses OpenAI models for text and images. Both are US companies and their models run in the US. A language model only gets the context a task needs, not your whole workspace.
Anthropic (Claude)
- Trains on your data
- No. The commercial terms forbid Anthropic to train models on what comes in through the API.
- Retention
- Deleted within 30 days. Longer if the law requires it, or up to 2 years if their abuse monitoring flags a request.
- Contract
- With Anthropic Ireland. Transfers to the US based on standard contractual clauses (SCCs).
OpenAI
- Trains on your data
- Not by default. OpenAI does not train on API data unless a customer opts in.
- Retention
- Up to 30 days, for abuse monitoring. Longer if the law requires it or if needed to protect their services.
- Contract
- With OpenAI Ireland. Transfers to the US based on standard contractual clauses (SCCs).
Sources, read on October 25, 2026: Anthropic's terms and retention periods, and the OpenAI API data policy.
For research, images, video and voice-overs we work with other specialised providers. They are listed with their purpose in article 5 of the privacy policy.
Retention
How long we keep things
A cleanup job enforces these periods automatically every night.
- Chat conversations on clients' websites
- Anonymised after 30 days without the visitor's consent. Deleted after 12 months with consent.
- Leads
- Deleted after 24 months.
- Inbox conversations
- Deleted 24 months after the last message.
- Video uploads that are not published
- Deleted after 24 hours.
If you cancel, you have 30 days to take your data with you. After that we delete it.
All periods, including those for Clyde's memory, Instagram messages and invoices, are in article 6 of the privacy policy.
Honest
What we do not promise
No zero data retention
We have no agreement with Anthropic or OpenAI under which they retain nothing. Their standard periods above apply.
The language model does not run in the EU
Servers, database and queues are in the EU, the language model is not. If you want the language model in the EU too, we run Clyde on Mistral on request.
Not everything stays in the EU
Vercel serves images and scripts through a global network, and some of the providers for research, images and video are in the US. Those transfers are covered by the safeguards in article 8 of the privacy policy.
Data processing agreement
Data processing agreement on request
We send a data processing agreement on request. Email us; you can also write with questions about your data or a data subject request.
privacy@future-marketing.ai