
Listen instead
A hospital network technology team monitors a sudden spike in application programming interface calls, tracing the traffic to a newly deployed clinical documentation assistant. Every time a clinician opens a chart, the external artificial intelligence tool makes dozens of individual database queries to pull past laboratory results, medication lists, and historical clinical notes. This high-frequency querying strains local server infrastructure and raises operational costs, demonstrating the friction at the intersection of electronic health records and external software.
To address this architectural challenge, Epic Systems has announced a shift in how it handles external artificial intelligence tools. On November 18, the company disclosed plans to allow authorized third-party artificial intelligence agents to retrieve electronic health record data directly through Model Context Protocol connections. Scheduled for deployment by the first quarter of 2027, the move represents a deliberate expansion of Epic's developer integration strategy, transitioning from a system that pushes clinical context outward to one that allows external tools to securely query the record on demand.
The Shift to Direct Database Queries
Currently, Epic's integration architecture relies on its Agent Factory Connectors. These connectors are designed to support outgoing integrations for specific artificial intelligence utilities, such as the digital assistants known as Emmie, Art, and Penny. Under this framework, Epic's internal systems initiate the handoff, sharing clinical context and tasks with external software in a controlled, one-way flow.
The update planned for early 2027 will introduce incoming Model Context Protocol, or MCP, connections. This standard will allow external, third-party agents to initiate contact and pull patient data directly from the electronic health record, provided they possess the necessary security credentials. Although Epic has established the timeline for early 2027, the company has not yet detailed the specific clinical workflows or administrative functions these incoming connections will support. For third-party software developers, the change means preparing their systems to query Epic databases using a standardized, open protocol rather than building bespoke interfaces for every hospital deployment.
Overcoming the Integration Bottleneck

Model Context Protocol is an open standard designed to establish secure, bidirectional connections between artificial intelligence models and data sources. By adopting this protocol, Epic is attempting to standardize the data pipelines that external developers must use to interact with clinical databases. This standardization could reduce the custom engineering work typically required when health systems deploy new software applications.
Managing these connections is a substantial technical challenge for hospital IT departments. Epic already supports a large integration footprint, with more than 3,700 connected applications across its ecosystem and over 1,000 technologies available through its Open.Epic platform. As hospitals adopt more external tools, IT leaders have raised concerns about rising application traffic and the associated costs of querying electronic health records.
To mitigate these infrastructure demands, Epic is introducing Fast Healthcare Interoperability Resources, or FHIR, notifications alongside expanded bulk-data processing capabilities. These features are designed to allow systems to process larger batches of information at once and receive automatic updates, reducing the need for constant, resource-intensive queries from individual artificial intelligence agents. This allows health systems to export larger datasets at scheduled intervals, optimizing network performance during peak clinical hours.
Sector Judgment: The Governance Frontier

The operational significance of this transition extends beyond technical compatibility. By adopting the Model Context Protocol, Epic is establishing the rules of engagement for how external artificial intelligence programs interact with clinical data. For clinical administrators, this change could accelerate the deployment of third-party tools that assist with documentation, billing, and clinical decision support.
However, the industry conversation is shifting from basic data access to governance. Health system technology leaders will need to evaluate which outside agents receive authorization to connect to their databases. The balance of power will depend heavily on who manages permissions, how clinical functionality is restricted, and how data access is priced. These operational decisions will determine the true utility of the incoming connections for daily hospital workflows.
Rather than relinquishing control, Epic is positioning itself as the central gatekeeper for third-party clinical AI. By establishing the integration standard, Epic ensures that even as health systems adopt external tools, those tools must operate within the structural and financial parameters set by the primary electronic health record vendor. CIOs must carefully calculate the total cost of ownership for these tools, weighing subscription fees against the infrastructure costs of running continuous database queries.
Source: The HealthTech Signal

