How These Examples Work
Each example demonstrates an AI agent reading from and writing to a Cascade Protocol Pod. The pod and the MCP server are local. Claude sends what it reads to Anthropic; use a local model to avoid that.
Reference Patient Pod
30 days of realistic data for a fictional patient, from the public conformance repository.
Two Access Methods
Use the Cascade MCP server with Claude Desktop, or invoke the Cascade CLI directly with Claude Code.
Prompt + Expected Output
Each example includes the exact prompt and a reference answer to evaluate the agent's response.
Medication Interaction Review
An AI agent reads a Cascade Protocol Pod, identifies a cross-provenance drug interaction between a prescribed ACE inhibitor and a self-reported potassium supplement, correlates it with an elevated lab value, and writes back a flagged observation.
Health Trend Analysis
An AI agent reads 30 days of wellness time-series data, identifies clinically relevant trends, ties confidence to provenance and sample density, and writes structured trend observations back to the Pod.
Visit Preparation
An AI agent reads a patient's entire Cascade Protocol Pod and generates a structured visit summary for an upcoming doctor appointment — distinguishing clinical-source from device-source data and flagging data gaps.