What You Should Know
- Ambulatory electronic health record (EHR) and practice management leader athenahealth launches athenaPatient MCP, an enterprise server integration built on the open Model Context Protocol (MCP).
- The infrastructure allows patients to connect their official electronic health record directly to Anthropic’s Claude, enabling conversational interpretation grounded in verified clinical history.
- The launch responds to changing patient behaviors highlighted by an August 2026 Pew Research Center survey, which reported that 34% of U.S. adults have used artificial intelligence assistants for health-related purposes, ranging from symptom exploration to diagnosing interpretations.
- The integration is live for Claude Pro and Claude Max subscribers across the United States, who can activate the athenaPatient MCP connector directly within their Claude account settings.
Patient-Controlled Context Brokering via MCP
Unlike static data exports or brittle screen-scraping plugins, the Model Context Protocol provides an open, standardized client-server convention transported over JSON-RPC. The athenaPatient MCP server functions as a governed middleware layer that structures EHR data before exposure to the LLM:
- Zero Static Pre-Loading: Rather than dumping the patient’s entire medical history into Claude’s training set or prompt window upfront, the athenaPatient MCP server serves as an on-demand data broker. Clinical records are queried dynamically only when the patient asks a specific question requiring historical context.
- Exhaustive Clinical Data Ingestion: Claude can reference discrete EHR data fields across the patient’s chart, including laboratory results, vital signs, active medication regimens, problem lists, immunization histories, encounter summaries, diagnostic imaging narratives, allergies, outpatient procedures, and longitudinal care plans.
- Direct Patient Credentialing: Security and consent are retained entirely by the consumer. Access requires patients to authenticate with their official athenahealth patient portal credentials, select specific clinical data scopes, and retain the unilateral ability to revoke model authorization at any time.
- Governance and Data Segregation: The interaction layer ensures that health information retrieved via the MCP endpoint remains restricted to the active session and is governed by strict zero-data-retention and non-training commitments, preventing personal health information (PHI) from entering foundation model pre-training corpora.
“The launch of athenaPatient MCP is an important milestone in responsibly integrating AI into the patient experience,” said Paul Brient, chief product and operations officer at athenahealth. “At its core, this is about helping patients get value from their own healthcare data. Anthropic has been a true innovator in advancing AI that is both powerful and responsible, and we are excited to work with them to leverage AI to meaningfully reduce complexity and help patients play a more active role in their care.”
