Data and security
Security and your data
Your brand is safe with us, and you stay in control of who uses it. People sign in with the work accounts they already have. Every request Claude makes is checked against who’s allowed in at that moment. Your brand data is stored in London, and no other customer can reach it.
How people connect
There are two ways in, and you choose which for each brand.
Work sign-in
Your team connects using the Google or Microsoft work account they already have. We never see their password. This uses OAuth 2.1 with PKCE, the current standard for connecting apps to Claude. You tell us your domain, and anyone with a verified email on it gets the brands you picked. We only trust emails that Google or Microsoft have verified. Public addresses like Gmail or Outlook can never be used to claim a domain, and each domain belongs to one account only.
Share links
For freelancers, partners or a quick pilot, you can create a link for one person. You name it, choose its brands, and decide whether it can save drafts or is view-only. You can give it an end date, and you can remove it at any time. We show the link once and store only a scrambled (hashed) copy, so it can’t be read back from our database.
Our sign-in pages are protected by Cloudflare Turnstile, which stops automated sign-up and sign-in attempts.
What Claude can and can’t do
Claude can read your brand and check work against it. It can’t approve anything. Editors’ saves from Claude go live straight away and can be undone. Everyone else’s go to the team for approval.
View-only links can’t save anything at all. Only the people you make owners or editors can approve, invite others or change settings, and they do that in the Creative in the Loop app, not through Claude.
Keeping brands separate
Each request from Claude covers one brand. A request covering more than one brand is refused, and work that mentions another brand gets a warning. That matters most for agencies looking after several clients.
If someone asks about a brand they can’t reach, they get the same answer as if it didn’t exist, so nothing gives away which brands we hold. Repeated guessing at links is blocked after a few attempts.
Where your data lives and who handles it
Your brand data and files are stored in London. The app runs in London and Washington, D.C. Our database can’t be reached from the public internet directly: every request goes through our app, which checks who’s asking first.
We use a small number of trusted companies to run the service:
| Company | What they do for us | Where |
|---|---|---|
| Supabase | Database, file storage and sign-in | London |
| Vercel | Hosts the app | London and Washington, D.C. |
| Anthropic | Reads the guidelines you upload and turns them into your brand | USA |
| Resend | Sending emails | USA |
| Cloudflare | Checks sign-ins come from real people | Global |
Anthropic’s commercial terms say it may not train its models on what we send it.
What we record
Each time Claude uses your brand, we record which connection it came through, who it was (or which share link), which brand, which tool and how long it took. That gives you an audit trail of who used what.
We don’t record what people typed into Claude or what Claude said back.
Removing access
We check access every time Claude asks, not only when someone first connects. Remove a person, remove a share link or switch a brand off, and it stops working on the very next request.
If someone leaves your company and their work account is closed, they can no longer sign in.
A brand owner can delete a brand from the app at any time. That permanently removes its uploaded files and everything else we hold for it.
Your data inside Claude
When Claude reads your brand, that content becomes part of the person’s Claude conversation. From there it’s covered by your organisation’s own Claude settings and Anthropic’s terms, not by us.
Questions or concerns
If your security or IT team has questions, or you think something’s wrong, email support@creativeintheloop.com. We’re happy to talk it through.