Electronic health record company athenahealth is opening up patient health records to outside AI assistants, starting with Anthropic’s Claude, through a new connection designed to give patients personalized answers based on their clinical data.
The new server solution, athenaPatient MCP, uses the Model Context Protocol to connect Claude directly with the athenaOne platform. Patients authenticate using their athenahealth patient portal credentials and decide what information Claude can access, according to the EHR company. Credentials are not shared with Claude.
Paul Brient, athenahealth’s chief product and operations officer, said information is encrypted in transit, patients control what information is shared, and they can revoke access at any time. “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,” he told Fierce Healthcare in an interview.
Brient noted that AI can meaningfully reduce complexity and help patients play a more active role in their care. “At its core, this is about helping patients get value from their own healthcare data,” he noted.
Athenahealth is committed to responsible AI integration, company executives said.
The company will limit athenaPatient MCP access to AI assistants that do not use patient information to train their models, he noted. “We also encourage patients to review the privacy and data-use policies of any service they choose to use. Healthcare data deserves a higher standard of stewardship than most information, and we’ve approached this work with that principle in mind,” Brient said.
Patients are turning to generative AI, like Claude and OpenAI’s ChatGPT, to interpret lab results, medication instructions and other medical information. More than a third of U.S. adults (34%) have used AI assistants for health-related purposes, including looking up symptoms to understanding a diagnosis, according to an August Pew Research Center survey. This trend reflects a growing expectation for immediate, conversational health guidance, but also introduces new governance challenges, athenahealth executives noted.
“Claude is the first connection we’ve enabled for athenaPatient MCP, but it’s not intended to be the only one,” Brient said. “Anthropic has been a strong launch collaborator because of its approach to responsible AI development. But our strategy isn’t built around a specific AI model. The AI landscape is moving incredibly quickly, with new models and tools coming to market seemingly every week. We designed athenaPatient MCP with that reality in mind and built it on a Model Context Protocol (MCP) that has become an industry norm for AI interoperability.”
Claude Pro and Max users in the U.S. can add the athenaPatient MCP connector within their Claude settings. Patients can then ask Claude questions about their health information using natural language. With a patient’s permission, Claude can reference information such as lab results, medications, vitals, health conditions, diagnostic reports, immunizations, visit history, allergies, procedures and care plans to generate personalized answers based on their clinical data, instead of generic or incomplete information, athenahealth executives said.
“Patients often have questions about their health records between visits, from what a lab result means to what to ask at their next appointment,” Syed Mohiuddin, head of healthcare at Anthropic, said in a statement. “athenahealth is giving patients a way to bring their own records into Claude when they choose to, so they can explore those questions and feel more confident and prepared in conversations with their care team.”
Athenahealth’s EHR is a cloud-based, open platform, Brient noted, and the company’s objective is to “create a secure way for patients to use their health information with the AI assistants they choose,” he said.
The company plans to expand access to other models, but Brient did not provide any details about additional integrations.
The company built athenaPatient MCP to connect to general-purpose AI assistants rather than build those generative AI capabilities entirely within athenaPatient.
“It’s not our job to decide which AI assistant patients should use. Patients are increasingly using AI tools in their daily lives, and I expect that trend to continue. If we’ve learned anything over the last few years, it’s that people will gravitate toward the experiences that work best for them. Healthcare shouldn’t be isolated from that reality,” Brient said.
“At its core, this is about giving patients access to their health information in the places where they want them—safely, of course. Our patient-facing connector is consistent with our broader interoperability capabilities. Consumers increasingly expect their personal data to integrate into the digital tools they use. Healthcare shouldn’t be an exception,” he said.
Some providers have raised concerns that as patients increasingly use AI chatbots to interpret health records and lab results, it will drive an uptick in follow-up messages to physicians. Brient noted that giving patients an easier way to connect their health records to AI tools helps patients better understand the information they have been given by their providers and have more informed questions.
“One of the realities of healthcare is that patients often leave an appointment with more information than they can reasonably absorb in the moment. Lab results often arrive after the patient has seen their doctor. Questions come up after the visit. Medication instructions get forgotten. That’s not a criticism of patients or clinicians. It’s just reality,” he said.
“When patients have a stronger understanding of their healthcare situation, conversations with their care teams tend to be more productive. From the clinician perspective, we were equally focused on what this should not do. We did not want to create additional work for clinicians. A recurring theme across our product strategy is reducing work while improving outcomes. The best healthcare technology helps patients become more engaged without creating more administrative burden for clinicians,” Brient noted.
Athenahealth provides EHR software, revenue cycle management and patient engagement solutions to small and independent provider practices. A year ago, the company unveiled new AI features for ambulatory providers as part of a larger upgrade to an “AI-native” athenaOne platform. This summer, the company rolled out more than 80 new AI features for RCM operations on its AI-native athenaOne platform.
Athenahealth executives describe the launch of athenaPatient MCP and the connection to Claude as just the first step, suggesting a broader strategy of allowing patients to use AI tools with their health records.
“The immediate focus is helping patients securely access and understand their health information through the AI assistants they choose to use. Looking further ahead, I think it’s reasonable to expect both the number of AI assistants and the capabilities available through them to expand. One of the reasons we’re excited about MCP is that it creates a standardized way to provide context to AI systems. Context is what makes AI particularly useful. Without context, AI can provide general information. With context, it can provide a more detailed, personalized set of information and is less likely to lead you astray,” Brient said.
Major EHR companies are moving forward with AI integrations to connect patient data with AI tools OpenAI’s ChatGPT for Healthcare now integrates with Epic’s EHR to enable authorized clinicians to access and summarize patient information from medical records within ChatGPT or directly in supported Epic workflows. That integration can now pull together clinical notes, lab results, medications and specialist documentation, the company said in a blog post.
Oracle’s AI-powered patient portal allows patients to chat directly with their electronic health record using a conversational interface built on OpenAI frontier models.
In healthcare, MCP creates a common framework for connecting AI agents to clinical systems and data.
“I think it’s still early, but this is an important trend to watch,” Brient said. “One of the challenges with AI is that intelligence without context isn’t particularly useful. In healthcare, context must be accurate, permissioned, secure, and current. That’s a much harder problem than simply connecting a chat to a data source.”
What makes MCP interesting, he noted, is that it provides a framework for connecting AI systems with the context they need while preserving appropriate controls and governance.
“Whether MCP ultimately becomes the interoperability layer for AI in healthcare remains to be seen, but I think the demand for this type of interoperability is only going to increase. Patients and providers are already using AI. The question is whether healthcare data can participate in those experiences safely and effectively,” he said, adding, “I feel good about where the industry is headed, and I believe standards like MCP can play an important role in getting us there. But we’re still in the early innings, and much work remains. Our goal is straightforward: reduce friction, maintain trust, and help patients and providers benefit from advances in AI without compromising security or control.”