Support, and how we use AI
Getting help
Write to support from the portal or by email. Tell us the app's name and,
if it is about a deploy, the deploy id (kaviris apps get shows the last
one).
Feedback and the wishlist
During the beta, tell us what breaks, confuses you or is missing:
- In the portal: the Feedback link at the top of every page. Pick
bug, idea or other and write what happened. The page and the app you were
on, and your browser's name, go with it unless you untick that. You get a
reference (
fb-…). - From your agent:
send_feedback(MCP) orPOST /v1/feedback, with the app, the deploy id and the refusal code it saw. An agent key needs no extra permission for this, only membership of its organisation; it may send 5 an hour and 20 a day. - The wishlist (portal, Wishlist): wishes from every tester. Add one, vote for the ones you want (one vote each, and you can take it back), and see what we decided: new, planned, in progress, done or won't do.
A person at Kaviris reads every report. What you write is never shown to other testers, except a wish you add yourself. Feedback is kept 24 months (longer only while it is linked to an open wish) and is deleted with your account.
Letting support look at your app
Support sees very little by default. To let it look closer at one case,
give a case grant: in the portal, or ask your agent ("give Kaviris
support access to recipes for case 1234 for 24 hours"), which calls
grant_support with the case number and a time.
- Time-limited: 1 to 72 hours, then it ends by itself.
- For one case and one app.
- Withdraw it any time (
revoke_support_grant, or in the portal). - You see every look: the list of grants (
list_support_grants) shows each thing support read under it, and when.
Support only reads. It can never change, deploy, restore or delete anything, and never sees your secret values or the rows in your database.
AI in support: what you should know
Answers from Kaviris support may be written by an AI system. We say so because you have a right to know when you are dealing with AI, and because it changes what you should share.
- Every reply drafted by AI says so. If you want a person to look at your case, say so in your message.
- Where it runs. The AI that reads your support case runs on models hosted in Europe (in Sweden, with a Swiss provider as back-up). Your case and your data are not sent to AI providers outside Europe; at most a summary without personal data is.
- What it can see. Without your permission: your app's state, refusal codes and whether your account is active. With your permission (a case grant you give for at most 72 hours, and can withdraw): the app's contract, deploys and refusals, build logs, error counts, and recent logs with emails, tokens and IP addresses masked.
- What it can never see. Your secret values (nothing in Kaviris can show them), your database contents, your uploaded files, other customers' data, and your card details.
- What it can't do. It only reads. It cannot change your app, deploy, restore or delete anything. Fixes go back to you or your coding agent.
- Everything is recorded. Each thing support looks at is logged and listed to you.
Your coding agent is your choice
Kaviris doesn't send your code to an AI. If you connect a coding agent (Claude Code, Cursor, …), that agent and its provider see what you show them, under your agreement with them. Kaviris keeps secrets out of that path: set them by reference.