Athenahealth Patients Can Now Query Their Medical Records Through Claude

Athenahealth, a healthcare provider that says it serves more than 20% of the US population, has launched a server that lets Anthropic's Claude read patients' medical records and answer questions about them. The tool, called AthenaPatient MCP, is available now — but only to patients who also pay for Claude Pro or Claude Max, subscriptions priced at $20 and $100 per month respectively. The company claims the arrangement is secure and that patient data cannot be used to train Anthropic's models. Those claims deserve scrutiny, and Athenahealth's own framing of what the tool does — and does not do — matters as much as the integration itself.

What Does the Integration Actually Do?

AthenaPatient MCP is a Model Context Protocol-compatible server. In plain terms, MCP is an open standard that lets AI applications connect to external data sources — here, Athenahealth's store of patient health records. Once a patient authorizes the connection, Claude can pull information from those records and use it to answer questions.

The access is not automatic, according to the announcement. Patients must sign in with their Athenahealth account credentials and explicitly grant Claude permission. Users can choose which categories of health information the model can see, and they can revoke access at any time, the press release states. Not all Athenahealth patients will have data shared with Claude, and not all Claude subscribers are required to opt in.

The scope of the data involved is broad. Once authorized, Claude takes into account lab results, vitals, medications, health conditions, immunizations, visit history, diagnostic reports, allergies, procedures, and care plans when answering patient questions.

That is essentially a patient's entire clinical file. Athenahealth frames the credential check as making the process more secure than a typical consumer pasting health details into a chatbot — a fair point, but one that rests entirely on the security of the MCP connection itself, which the company describes only in general terms.

Takeaway: patients must actively hand over their credentials, and can take them back, but the data in play is comprehensive.

Is This Medical Advice? No — And Athenahealth Is Explicit About It

The most carefully worded part of the announcement concerns what Claude will not do. The partnership does not provide medical advice, Sam Lambson, vice president of product management, data & ecosystem at Athenahealth, told ZDNET in an email. It should not replace a clinician's professional judgment.

"Patients should contact a qualified healthcare professional with questions about diagnosis, treatment, or changes to their care," Lambson said.

The practical effect is a narrow, informational use case. A patient can ask Claude something factual — "What was my most recent red blood cell count?" is the example ZDNET cites — but should not expect the model to explain what an abnormally high or low count means for their health. The tool appears designed to help patients better understand information health professionals have already given them, not to interpret it.

That boundary is sensible. It is also one that chatbots, by their nature, do not always respect — a model with full record access may volunteer explanations regardless of disclaimers, and patients may act on them. The press release does not address how Athenahealth or Anthropic police that line in practice.

Takeaway: the tool answers questions about records, not questions about health, at least on paper.

Why Health Data Is the Prize

The context for this deal is a broader race between big tech and big healthcare. Earlier this year, Microsoft partnered with Mayo Clinic to develop an AI model for healthcare. Boston Children's Hospital announced a partnership with OpenAI to build an enterprise AI system spanning research, clinical, and administrative teams.

Health data is extremely valuable, and technology companies are increasingly interested in partnering with providers to obtain it. Much of the optimistic "AI could cure cancer" messaging depends on large companies accumulating enough health data to train models and link disease data to patient data. That is not necessarily the motive here — Athenahealth says its data cannot be used to train Anthropic's models — but the industry-wide pattern is unmistakable, and patients reading a press release rarely see the contractual fine print behind such assurances.

A more modest explanation, and a plausible one, is administrative relief. An AI that can field routine patient queries — questions about test results, for instance — could reduce the between-visit admin work that doctors currently handle themselves. If Claude answers the "what was my count" questions, clinicians get time back for the ones that matter.

Takeaway: the same data that trains future medical AI also cuts doctors' paperwork today, and both incentives point toward deals like this one.

What Do Patients Actually Think?

Demand exists. A recent Pew Research Center survey found that 34% of US adults have talked with a chatbot for medical advice. That is a substantial share of the population already using general-purpose AI on health questions, typically without any authenticated connection to their actual records. Athenahealth's credential-based approach is, at minimum, a more structured version of behavior that is already widespread.

Confidence is another matter. The same Pew research found Americans wary of AI in clinical settings: the largest shares of respondents said they still prefer that a human oversee AI when it is used in a clinical context. Athenahealth and Anthropic have not released evidence that record-connected chatbots improve patient outcomes; whether partnerships like this one will remains, by ZDNET's assessment, to be seen.

The paywall adds another wrinkle. Because the feature requires Claude Pro or Claude Max, the patients best positioned to query their own records through AI are those willing to spend $20 to $100 per month on a chatbot subscription. A tool pitched as helping patients understand their care is, for now, a premium product layered on top of the care itself.

Takeaway: a third of US adults already ask chatbots about health, but most want a human still in charge — and this feature is reserved for paying subscribers.

What Comes Next

Athenahealth has not said whether AthenaPatient MCP will expand beyond Claude Pro and Max users, whether other AI clients could connect through the MCP server, or what auditing exists to verify the no-training commitment. The company has also not published data on accuracy — whether Claude reliably retrieves the correct lab value or medication from a record, and what happens when it does not, is the kind of question that only surfaces once real patients use the tool at scale.

For now, the integration stands as one of the most direct connections yet between a mainstream AI assistant and the medical records of a large share of the US population. Its stated limits — no advice, no training, opt-in only — are clear. How those limits hold up in practice is the story to watch.