athenahealth lets patients connect their records to Claude through a new MCP link
athenahealth is opening a direct channel between patient records and outside AI assistants, beginning with Anthropic’s Claude. The new connector, athenaPatient MCP, is designed to let patients ask questions about their own clinical data in natural language, while keeping authentication and access control inside the athenahealth patient portal.
The move pushes patient-facing AI one step closer to routine care. Instead of forcing patients to navigate a portal, decode lab panels, or search generic web results, the system is built to return answers grounded in the record itself — when the patient chooses to share it.
Patients authenticate with their athenahealth portal credentials, select which types of information Claude may access, and can revoke that access at any time. Credentials are not shared with Claude. Data is encrypted in transit, and athenahealth says the connector will only be available to AI assistants that do not use patient information to train their models.
Permissioned context, not raw data export
The architecture matters because healthcare data is not useful in the abstract; it becomes useful when it is paired with context. A lab value can look alarming without reference ranges, prior trends, active medications, or a recent visit note. A medication instruction can look simple until a patient needs to know whether it still applies after a dose change or specialist referral.
athenaPatient MCP uses the Model Context Protocol to connect Claude directly with the athenaOne platform. In practice, that means the AI assistant can draw on information such as lab results, medications, vitals, health conditions, diagnostic reports, immunizations, visit history, allergies, procedures, and care plans. The goal is not to replace the clinician-patient conversation. It is to make the information that already exists in the record more intelligible between visits, when confusion often begins.
Paul Brient, athenahealth’s chief product and operations officer, said the company built the connector around patient control, encryption, and revocable access. Once a patient approves the data types to share, the interaction moves between the patient and Claude.
Patients are already using generative AI for health questions at scale. A recent Pew Research Center survey found that 34% of U.S. adults have used AI assistants for health-related purposes, from symptom checks to understanding diagnoses. That behavior is changing expectations. People now want answers that are immediate, conversational, and personalized — and they are increasingly willing to bring their records into the exchange.
Follow-up questions may get better, and more specific
A single message from a patient can now begin with a lab panel and end with a more informed visit agenda. That may be the most practical consequence of the launch.
Brient framed the connector as a way to help patients become more active participants in care without creating new burden for clinicians. He pointed to a familiar pattern: patients leave visits with more information than they can absorb, lab results arrive after the appointment ends, and medication instructions are often forgotten before the next encounter. The technology is meant to fill that gap by helping patients interpret what they already received, not by sending them back to the clinician for every basic question.
That design goal is important because some provider groups have warned that patient-facing AI could generate more inbox traffic. athenahealth is arguing the opposite may happen if the tool improves comprehension first. When patients can sort through their own records with an AI assistant, the questions that reach the care team may be more focused and useful.
Anthropic’s head of healthcare, Syed Mohiuddin, said Claude can help patients explore questions that arise between visits and feel more prepared in conversations with their care team. For patients, that could mean fewer blind spots around a new diagnosis, a medication change, or an abnormal result that appears after the office visit has ended.
Responsible AI rules are central to the launch
Healthcare AI lives or dies on trust, and trust here depends on governance as much as on model quality. athenahealth is limiting the connector to AI systems that do not train on patient information, and the company is advising patients to review each service’s privacy and data-use policies before connecting anything sensitive.
Brient said healthcare data deserves a higher standard of stewardship than most information. That is not marketing language; it is the core safety issue. Clinical data are not just personal — they can reveal diagnoses, procedures, family history, medication patterns, and future care needs. Once that information leaves a medical workflow, control can become murky unless the rules are explicit from the start.
The company also emphasized that Claude is the first connection, not the only one. athenahealth plans to expand access to other models, and the connector was built around MCP specifically because the standard can support interoperability across different AI systems. Brient said the strategy is not centered on a single model or vendor. It is centered on a standardized way to deliver permissioned context to AI tools patients already use.
EHR vendors are racing to define the patient-AI interface
athenahealth is not moving alone. Epic has begun integrating ChatGPT for Healthcare so authorized clinicians can access and summarize patient information from medical records within ChatGPT or supported Epic workflows. Oracle has also introduced an AI-powered patient portal built on OpenAI frontier models. Across the sector, major EHR players are trying to decide where conversational AI should live: inside the electronic record, in a standalone assistant, or at the junction of both.
athenahealth is choosing the connector model. Instead of building the full generative AI experience inside athenaPatient, it is creating a secure bridge to outside assistants. The company says that approach better reflects how patients already use digital tools in daily life. They will gravitate toward the experience that works best for them, Brient said, and healthcare should not be isolated from that expectation.
The unresolved operational question is not whether patients will use AI. They already are. The harder question is whether healthcare systems can let those tools work with live clinical data without weakening security, consent, or data stewardship.
For now, athenaPatient MCP is a first step toward that answer. The connector is live with Claude for U.S. Claude Pro and Max users, and athenahealth plans to broaden access over time. What remains unproven is how widely patients will adopt the feature, how consistently they will interpret the responses, and how much downstream workload it creates or reduces for practices managing the messages that follow.