Most AI health assistants answer questions about medicine in general. athenahealth went a step further this week, launching athenaPatient MCP, a connector that lets patients authorize Anthropic’s Claude to read their own athenaOne charts.
Patients sign in with portal credentials, pick which categories to share, and can withdraw permission any time. Those credentials never reach Claude, and data is encrypted in transit. athenahealth says the connector stays closed to any assistant trained on patient records at all.
What the assistant can actually see
Once connected, the chart opens up: lab values, prescriptions, vital signs, diagnoses, imaging and pathology reports, shot records, clinic visits, medication allergies and care plans. Patients ask in plain language, or plan what to raise at a coming appointment.
Paul Brient, chief product and operations officer, framed the tool as help for patients making sense of records between visits without adding clinician work. Anyone on Claude’s paid Pro or Max tier in the US turns it on through their own account settings.
The protocol is the interesting part
MCP matters more than the first integration. MCP is becoming the standard way AI systems reach outside data, and an EHR vendor adopting it creates a template other assistants can follow. Brient said Claude is first, with other models planned.
athenahealth serves mostly small and independent practices, where IT budgets are thin and proprietary AI is out of reach. A 2026 Pew survey finding that 34% of US adults had already asked an assistant about symptoms makes that demand hard to ignore.
The field is crowded. Epic has linked records to OpenAI’s healthcare offering, and Oracle has shipped a conversational portal. The differentiator here is granular consent, and athenaOne’s 80-plus billing AI features suggest the company wants patients and revenue teams on one platform.
