Staff portal
Sign in to work your practice queue — callbacks, approvals, conversations, and reports for your practice only.
Work
One list of everything waiting on this practice — urgent first, then whatever is overdue, then whatever is due soonest. Each row says one thing to do about it and who it belongs to. Rows you cannot act on still say what has to happen and who has to do it. Arrow keys move; Enter does the selected row's next step.
Callbacks
Patients the agent could not finish with — each one is waiting for someone to ring them back — and messages logged, including those that never reached the EMR. This is not the approval queue: nothing here needs reviewing, it needs doing.
Conversations
Patient conversations, newest activity first. Opening one shows the full record — the transcript with every tool call and refusal beside it. Each read is written to the audit log.
Memory
Things the AI employees propose to remember about your practice. Nothing takes effect until you promote it; promoted entries shape every future draft, so read them the way you would read a new standing instruction to staff.
Crisis
Crisis obligations — standing duties to reach a clinician, opened before the patient saw a reply. Read-only: a duty is discharged only when a clinician acknowledges the page from their own phone, never from this screen.
Reports
Live health — open callbacks, crisis pages, dropped Clara/Felix conversations — and the last 24 hours as a digest. Counts and reasons only; full transcripts stay on Conversations.
Learning
Patterns found in what actually happened — what failed, what a colleague corrected, and how often — each with the evidence behind it. Every row is a PROPOSAL: approving one creates a candidate for you to promote on Memory, or a packet for engineering. Nothing here changes what the assistant says until it has been qualified and released separately.
Activity
Human actions and security refusals from the tamper-evident audit log (most recent first).