athenaPatient MCP Brings Patient Records Into Claude

Estimated reading time: 7 minutes · Last updated: 2 November 2026

Athenahealth, the electronic health record (EHR) company, now lets patients send their own clinical records to outside AI assistants, starting with Anthropic's Claude. The connection is called athenaPatient MCP. A patient signs in with their athenahealth portal login, chooses which parts of their chart Claude may read, and then asks about it in plain language. Claude never receives the login. U.S. subscribers on Claude's Pro and Max plans switch it on in Claude's settings. Athenahealth says it will only connect assistants that keep patient information out of model training. The company's wager is simple. Patients already ask chatbots about their health, so their records should be able to follow them there.

It's not our job to decide which AI assistant patients should use.

Paul Brient, chief product and operations officer, athenahealth

How athenaPatient MCP Links a Patient Portal to Claude

The Model Context Protocol, or MCP, is a shared standard that lets an AI assistant draw context from an outside system. Athenahealth built athenaPatient MCP as a server on that standard. It wires Claude directly into athenaOne, the company's cloud-based EHR platform. A patient proves their identity with their athenahealth patient portal login, and that login stays with athenahealth.

After sign-in, the patient sets the scope. Paul Brient, athenahealth's chief product and operations officer, said the data is encrypted while it moves. The patient picks what to share and can withdraw access whenever they like. Brient described what happens next in an interview, as first reported by Fierce Healthcare: "Once the connection is established and patients approve what types of data they would like to share, the rest is between Claude and the patient."

Setup happens inside Claude. U.S. users on the Pro and Max plans add the connector in their settings. With permission, Claude can draw on several kinds of record: allergies, immunizations, medications, vitals, and lab results and diagnostic reports. It can also see health conditions, procedures, visit history and care plans. Athenahealth executives say this lets the assistant answer from a person's actual chart rather than from generic or partial information.

Why athenahealth Bars Assistants That Train on Patient Data

The connector comes with a condition attached. Brient said athenahealth will grant athenaPatient MCP access only to assistants whose makers keep patient information out of model training. He also urged patients to read the privacy and data-use terms of whatever service they pick. "Healthcare data deserves a higher standard of stewardship than most information," Brient said.

That leaves an open point. In Brient's own description, the relationship after approval belongs to Claude and the patient. So once records flow across, the protections that apply depend largely on Anthropic's policies rather than athenahealth's. The training restriction is athenahealth's gate, and the patient's reading of the assistant's terms is the second one.

Anthropic is presenting the tool as preparation for appointments. Syed Mohiuddin, head of healthcare at Anthropic, said in a statement that patients often have questions between visits, such as what a lab result means or what to raise next time. He said the link should help them "feel more confident and prepared in conversations with their care team." Brient called Anthropic a strong launch collaborator because of its approach to responsible AI. "Claude is the first connection we've enabled for athenaPatient MCP, but it's not intended to be the only one," he said.

Will AI Chart Questions Add to Doctors' Inboxes?

Patients are already using generative AI to read their own medical information. Tools such as Claude and OpenAI's ChatGPT get asked about lab results and medication instructions. Some 34% of American adults had turned to AI assistants for health reasons, from checking symptoms to making sense of a diagnosis, an August Pew Research Center survey found. Athenahealth executives say the habit raises expectations of instant, conversational answers. They say it also creates new governance problems.

Some providers fear a side effect. They expect that as more patients run records past chatbots, physicians will field more follow-up messages. Brient argues the reverse. He says a patient who better understands what their doctor already gave them comes back with sharper questions. He pointed out that lab results often arrive after the visit and that instructions get forgotten. "That's not a criticism of patients or clinicians. It's just reality," he said.

Brient said the design work focused just as hard on what the tool should not do. "We did not want to create additional work for clinicians," he said. For now, that is a stated intention. Athenahealth has published no figures on message volumes, so the provider concern and Brient's answer both remain untested.

How athenahealth's Bet Differs From Epic and Oracle

Other EHR vendors are linking records to AI too. OpenAI's ChatGPT for Healthcare now works with Epic's EHR. Authorized clinicians can retrieve and summarise patient information in ChatGPT or in supported Epic workflows. According to a company blog post, the tool can now draw together a patient's medication list, lab findings, notes written by clinicians and paperwork from specialists. Oracle's AI patient portal offers a chat interface to the health record, built on OpenAI frontier models.

Athenahealth chose a different route. It did not build generative AI features entirely inside athenaPatient. Instead, it connects to general-purpose assistants that patients already use. "It's not our job to decide which AI assistant patients should use," Brient said. He added that he would not tie the strategy to one model. He gave no detail on which integrations come next.

Brient also hedged on the standard itself. He said it is unclear whether MCP will become the interoperability layer for AI in healthcare. He said in healthcare, context has to be accurate, permissioned, secure and current. The company sells EHR, revenue cycle management (RCM) and patient engagement software to small and independent practices. A year ago it unveiled AI features as part of an upgrade to an "AI-native" athenaOne. This summer it added more than 80 AI features for RCM work.

What could help or hold back athenaPatient MCP

The bottom line

athenaPatient MCP is a narrow launch with a broad aim. For now it serves one assistant, U.S. Claude Pro and Max users, and whatever parts of a chart each patient approves. Brient presents it as a first step toward letting patients take their records to whichever AI tool they prefer. It is built on a standard he hopes will spread but will not call settled. The open questions are practical. Athenahealth has not said which assistant comes next, or when. Its promise of no extra clinician work has not been tested against real inbox volumes. And once a patient approves the link, much of the data's handling sits with Claude. That makes the assistant's privacy terms as important as athenahealth's own controls.

What to watch

• Watch for athenahealth to name the next AI assistant cleared for athenaPatient MCP; Brient said more are planned but gave no names and no date.
• Watch whether practices on athenaOne see a change in patient follow-up messages once the Claude connector is in use; no timeline for any data has been given.
• Watch whether MCP gains ground as the standard for connecting AI to clinical systems, a question Brient said is still open; no date has been set.