Security
It reads your mail. That is why it has clear limits.
What draften may do, what it may not, and what it keeps.
Never sends on its own
It only writes drafts into your mailbox. Only a person can send them.
Starts in trial mode
A new assistant writes only into draften. You let it into your mailbox when you are satisfied.
Nothing applies without you
Lessons from your edits and knowledge corrections apply only once approved.
Access encrypted
Passwords, keys and secrets are stored encrypted and never shown again.
Roles and change history
You decide who may change what. Changes to settings go into a history.
Never writes to your systems
It only reads from your CRM. We help you give it read-only access.
Mailbox access
We do not hide what it needs.
To create drafts and label messages, draften needs read and write access to the mailbox.
What it does in the mailbox
What it does not do in the mailbox
It sends nothing, moves nothing, never deletes your messages and never marks them as read. In trial mode it writes nothing into the mailbox at all. Neither Microsoft nor Google offers a permission for drafts alone, so the permission granted allows more; draften uses it only to read, to create drafts and to add labels.
Data and models
Where it runs and whom it trusts with what.
Where it runs
In our cloud, on Cloud4medical s.r.o.’s own infrastructure in the Czech Republic (EU). Or entirely with you, on your own server.
What it keeps
In its everyday work it does not store the text of incoming emails. It keeps the sender and subject of messages, the drafts it wrote and how people edited and sent them – for 90 days at most, then it deletes them. On your own server you set that time yourself.
Who reads it on its behalf
The model you choose. Neither Anthropic nor OpenAI trains on data sent through their APIs, and both keep it for 30 days at most. With a local model, your mail goes to no one else.
Need a data processing agreement?
It is part of our terms of use. If you need your own wording, write to us.