Patients are already pasting lab results into chatbots. The question has never really been whether they will ask AI about their health, but whether the AI will see anything more than a screenshot and a guess. athenahealth is betting that a direct, patient-authorized connection to the actual record is a better answer.
On November 2, 2026, the company announced athenaPatient MCP, a server that lets patients connect their athenahealth records to Anthropic’s Claude. It is available starting today.
Why patients are asking for this
The demand is easy to see. An August 2026 Pew Research Center survey found that 34% of U.S. adults have used AI assistants for health purposes, from checking symptoms to making sense of a diagnosis. Most of that use happens with no access to the person’s real data. The answers tend to be generic, or they rest on whatever the patient remembers to type.
That gap matters clinically. A person who misreads a lab value or misremembers a medication dose can walk into an appointment with the wrong idea. Connecting the assistant to the source record is one way to narrow that risk, though it does not remove it.
How the connection works
The tool uses the Model Context Protocol, an open standard that gives AI models defined, secure points of access to structured data in authorized systems. The athenaPatient MCP server links Claude to athenahealth’s platform. When a patient asks, Claude can retrieve record information through a secure connection that requires sign-in with the patient’s athenahealth credentials.
Claude Pro and Max users in the U.S. can add the connector in their Claude settings. After that, they can ask questions in plain language. Claude can reference the following data:
- Lab results and diagnostic reports
- Vitals
- Medications and allergies
- Health conditions and procedures
- Immunizations
- Visit history
- Care plans
The company says patients choose which information to make available and can revoke access at any time. Pricing for the connector itself was not disclosed. Access depends on a paid Claude plan.
What the companies say
Paul Brient, athenahealth’s chief product and operations officer, called the launch “an important milestone in responsibly integrating AI into the patient experience,” and said the goal is to help patients get value from their own data. Syed Mohiuddin, Anthropic’s head of healthcare, pointed to the questions that come up between visits, such as what a result means or what to ask next time. The aim, he said, is to help patients feel “more confident and prepared” when they talk with their care team.
That framing is worth noting. Neither company is pitching this as a replacement for clinical judgment. It is positioned as preparation for the visit, not a substitute for it.
What clinicians and entrepreneurs should watch
For clinicians, the practical effect may be patients who arrive with sharper, more specific questions. That can be useful. It could also mean more portal messages and more time spent correcting an AI’s misreading of a result. The release offers no data on accuracy, error rates, or how Claude handles ambiguous findings, and those details will matter.
The release also leaves open how patient data is treated once it reaches the AI provider. Patient-directed access to records is generally not governed the way a provider’s own systems are, so the consent and revocation controls carry real weight. Patients should read the terms before connecting.
For health entrepreneurs, the signal is the architecture. An EHR vendor is exposing patient data through a standard connector rather than building its own chatbot. If that approach spreads, interoperability work may shift toward making records easy for any AI tool to consume, with the patient holding the keys.
athenahealth calls this launch “just the beginning” and describes it as a step toward patients using AI to understand their records and prepare for care decisions. Whether it improves understanding or simply adds a new source of confident-sounding errors will depend on evidence the companies have yet to publish.