Hosting
The app, API and database run on dedicated servers in Germany. Uploaded files, encrypted backups and email delivery run on AWS in Ireland. Data stays in the European Union.
Servers are configured as code and hardened: key-only access on a non-standard port, no root login, a firewall that only opens the web ports, automatic security updates, and separate least-privilege accounts for administration, deployment and read-only debugging.
Encryption
- All traffic uses TLS.
- The tokens that connect your tools are encrypted at rest with a key held only by the Clearkeep servers.
- Nightly database backups are encrypted before they leave the server.
Access you decide
A connected account belongs to the person who connected it. Nobody else can use it until its owner shares it with people or groups, and for calendars, calendar by calendar.
Each account lists its actions one by one, reads and writes apart. Irreversible writes, such as deleting an event, start off on every new connection. Claude, ChatGPT or any other assistant plugged into Clearkeep only ever acts through an action that is on, for a person it is shared with.
Audit log
Every call an assistant makes is recorded with who it acted for, the action, the calendars it touched, the assistant, and whether it was allowed. Changes to connections, groups and access are recorded too. Workspace admins see the whole log in the app.
Sign-in
There are no passwords. People sign in with a one-time code sent to their work email, valid for a few minutes. Each workspace has its own address, and membership is managed by its admins.
Retention and deletion
Calendar events are fetched from Google on each call and never stored. For a connection, Clearkeep keeps its encrypted tokens, its list of calendars and the grants on them; the audit log keeps which calendars and events each call touched. Removing a connection deletes its tokens, calendars and grants at once; Clearkeep's access at Google is revoked with the account's last Google connection. Backups rotate out within 6 months. The full schedule is in the privacy policy.
AI models
Clearkeep runs no AI model of its own and never uses your data to train one. The assistant you plug in reads only what its grants allow, under your own agreement with its provider. A model provider Clearkeep adds later is listed as a subprocessor first and is barred from training on your data or keeping it.
Subprocessors
The current list, with what each provider does and where, is in the privacy policy.
Report a vulnerability
Write to support@clearkeep.ai with the details and how to reproduce it. We answer within two working days and credit you once it is fixed, if you wish. Please give us time to fix it before making it public, and do not access data that is not yours.