Restrict agents to patients associated with the active encounter, assignment, or approved workflow.
The patient relationship, not employment, grants access — the agent reaches the chart in front of it and no others.
Limit agents to the right patient, purpose, care team, and clinical action.
Healthcare agents may assist with scheduling, clinical documentation, patient communication, coding, prior authorization, and care coordination. Access cannot be based only on whether someone works for the organization—it has to follow the role and the patient relationship of the person the agent is acting for.
The chart in the next room belongs to someone else’s patient. This request never reaches it.
Example appointments. The panels, purposes of use and signing rules are the ones you set.
The patient relationship, not employment, grants access — the agent reaches the chart in front of it and no others.
Each is its own grant held by different staff and different agents, so a billing agent and a documentation agent reach different parts of the same record.
Segment-level scoping keeps sensitive portions of the chart out of workflows that have no need for them.
None of them is an action an agent holds, so anything clinical that reaches a patient or a chart is a licensed clinician’s own act.
Scheduling and billing agents work from demographic and coverage data without a path into clinical records.
Every action carries an accounting of who and what was involved — disclosure-ready by construction.
Explore the context attached to every request.
documentation-agent has a defined purpose and a bounded set of tools.
The request acts for treating clinician, using that person’s permissions.
The policy applies to the requested action within Active encounter, before the tool executes.
Documentation agents draft from the active encounter, and notes enter the chart under the clinician’s review and signature.
Intake agents collect and verify one patient’s information at a time, limited to that registration.
Scheduling runs on demographics and availability, with no clinical record access attached.
Billing and coding agents read the minimum necessary for the claim, with payment-purpose access recorded as such.
Authorization agents assemble the request from the patient’s own records; submitting to the payer is not an action they hold, so staff send it.
Coordination agents work across the care team for their assigned patients, within each member’s role-based scope.
Research use runs under its own purpose and its own grant, separated from treatment and operations access.
No. Fabriq drops in between agents and the clinical and administrative systems you already run. The EHR stays the system of record; every agent request against it is capped at the acting user’s own role and the grant the agent holds.
No. Administrative and clinical access are separate scopes; a scheduling or billing agent works from demographic and coverage data and has no path into clinical records, even when both connect to the same systems.
Immediately — sessions die on revocation rather than at token expiry, and a disabled agent’s next call fails closed.
Yes — that is exactly how it resolves. The same documentation agent reads differently for a treating clinician, a scheduler, and a biller, because reach is resolved per person and per purpose, down to segments of the record.
Every agent action carries the person who asked, the agent that acted, what was touched, and the verdict — one trail you can query when access questions come up.
In Fabriq’s vault, injected per call — no shared passwords in bot configs, automatic refresh, and instant cutoff when a connection is removed.