ExerScript

A physical-activity prescription agent demonstrating how healthcare-specific tools can work together through MCP and agent-to-agent orchestration.

Problem

Physical-activity prescription needs structured domain context and a workflow that keeps the output understandable and reviewable.

Users

Healthcare professionals exploring structured support for physical-activity prescriptions.

Context

An independent healthcare AI prototype, not a deployed clinical system.

Role

Designed and built the prototype, its MCP server and its agent-to-agent orchestration.

Requirements

• Support a physical-activity prescription workflow
• Expose healthcare-specific capabilities through an MCP server
• Coordinate specialised components through agent-to-agent orchestration
• Keep the result available for human review

Approach

Separated domain capabilities into tools, then coordinated them through an agent workflow instead of relying on one undifferentiated prompt.

Workflow

Healthcare context enters the agent workflow, specialised tools contribute through MCP, and the orchestrated result is returned for review.

Technology

Model Context Protocol (MCP), Agent-to-agent orchestration, AI agents

Safety and risk considerations

The prototype is not validated for clinical use and is not presented as an autonomous diagnosis or treatment system. Any clinical use would require formal evaluation, governance and human oversight.

Outcome

Produced a working healthcare AI prototype that demonstrates physical-activity prescription support using an MCP server and agent-to-agent orchestration.

Metrics

Not measured

Lessons

Healthcare agent design is clearer when domain capabilities, orchestration and human review are explicit parts of the workflow.

Next steps

Define a specific intended-use case, test with representative users and data, and agree clinical, privacy and quality acceptance criteria before a real-world pilot.