Standardized Healthcare Agent Remote Protocol (SHARP)

SHARP-on-MCP

SHARP-on-MCP: A foundational framework enabling independent, pluggable remote Model Context Protocol (MCP) servers to provide specialized healthcare services that can leverage clinical data from FHIR and other diverse healthcare data sources.

The rapidly evolving landscape of AI agents in healthcare necessitates a standardized, interoperable protocol for robust communication with external, server-based healthcare services. Currently, the absence of such a standard impedes the seamless integration of AI applications for critical functions like data summarization, extraction, clinical guidance, and workflow automation across various healthcare settings. The Standardized Healthcare Agent Remote Protocol (SHARP) is an open-source specification designed to address this critical interoperability gap, enabling consistent and reliable access to diverse healthcare data sources, including FHIR.

SHARP-on-MCP represents the initial realization of the SHARP specification. It is strategically built upon the Model Context Protocol (MCP), an open standard that facilitates the interaction of AI agents with external tools and data. While MCP supports both local and remote server implementations, SHARP-on-MCP specifically focuses on defining and standardizing the communication patterns for remote MCP servers. This focus is paramount to enabling the development of truly reusable and pluggable healthcare agent services that can operate independently of the client or agent's local environment. SHARP-on-MCP is engineered to provide a robust, foundational layer for the development of pluggable, composable, and interoperable agent-facing services, leveraging existing MCP specifications. While initially focused on MCP, the SHARP specification is designed for extensibility and may evolve to support additional communication protocols in the future.

Our overarching objective is to achieve for remote MCP server-like services what SMART on FHIR accomplished for Electronic Health Record (EHR) applications: to establish a universal, reusable architectural paradigm. This paradigm will enable the independent development, hosting, and seamless integration of these remote MCP servers with any FHIR server or other disparate healthcare systems and databases, thereby significantly enhancing interoperability and accelerating innovation in healthcare AI.