Permissions and safety
What an AI agent connected to Daybook can see and do, what it never sees, what waits for a person, and how its work is recorded.
An agent connected to Daybook is a person's assistant or an account's integration, never a new kind of user. It can do what its credential can, no more, and everything it does is on the record under that credential.
It acts as you
An assistant token reads as the person who made it: only the practices, pages and actions their role opens in the Daybook app. A biller's Claude sees a biller's worklist; it cannot reach team settings the biller cannot. A secret key acts as the account, as an API request does.
What it never sees
- A patient's birth date, member ids or full identity. Patients are shown by display name.
- A raw 835, 277CA or 999 file. Tools return what Daybook read from them.
- Another workspace's data. A token names its workspace, and an id from another answers as not found.
- Assistant tokens. A token cannot reach the page that makes them.
What waits for a person
Portal changes and posting approvals wait for a person in the Daybook app. An agent can ask; it can't approve its own request. Writes that only record, such as a claim note or a work record, go through at once. Claude and Cursor ask before every tool that acts or writes, unless you allow it always.
On the record
- Notes and work records show on the claim and the work item under your name.
- Calls, checks and portal tasks an agent starts are in the Daybook app with everything else, and send the same webhooks.
- Making and revoking a token is in Settings → Audit log. Profile → AI assistants shows when each token was last used.
- A write carries a request id, so a client that retries records it once.
Revoking access
Revoke a token in Profile → AI assistants; the app using it stops at once. Revoke a secret key in Developers → API keys.