Bringing AI-Powered Natural Language Access to a Radiology Platform with MCP
Integrated Anthropic Claude with the Model Context Protocol (MCP) into a 14-application radiology platform, enabling radiologists to query live clinical, financial, and operational data through natural language — reducing navigation overhead and accelerating decision-making.
Company
Private Limited
Domain
Radiology & Diagnostic Imaging
Service
AI Engineering & Platform Integration
Technology Stack
Problem Statement
Navigating Complex Radiology Data Across 14 Applications Was Slow and Manual
Radian's radiology platform comprised 14 independently deployable applications spanning clinical reporting, scheduling, financial settlement, and administration. Radiologists, technologists, and administrators had to navigate multiple dashboards, menus, and application interfaces to access worklists, patient history, visit reports, invoices, and operational data. In a fast-paced radiology environment where every minute matters, this navigation overhead slowed decision-making and reduced productivity. The organization needed an intelligent, conversational interface that could provide instant access to live platform data without requiring users to switch between applications or learn complex navigation paths.
Key challenges included:
- Users navigating multiple applications and dashboards to find routine information.
- No unified query interface across clinical, financial, and administrative data.
- Radiologists losing time switching between worklists, patient history, and reporting screens.
- Administrators manually checking RIS assignments, user roles, and group configurations.
- Financial teams navigating separate applications for invoices, payments, and settlement data.
- No natural language interface for querying live platform data.
- Role-based data access requirements across 7 distinct user roles.
The solution included:
- Implemented a multi-service MCP architecture with dedicated MCP servers for Connect, Accounts, and Calendar domains.
- Integrated Anthropic Claude Haiku as the AI engine with a tool-use loop supporting up to 10 iterative tool calls per query.
- Exposed 60+ MCP tools covering worklists, patient history, visit reports, dashboards, invoices, schedules, and user management.
- Built 4-layer authentication chain from browser JWT through to MCP server tool execution.
- Implemented role-based tool authorization at both advisory (system prompt) and enforcement (hard block) levels across 7 user roles.
- Designed async session persistence to MongoDB for conversation history and audit trails.
- Built a React-based chat widget with markdown rendering, session management, and role-aware interactions.
Solution
Building an AI-Powered Conversational Interface with MCP
Peramal engineered an AI assistant powered by Anthropic Claude and the Model Context Protocol (MCP), giving users natural language access to live data across all 14 Radian applications through a single conversational interface.
Impact & Benefits
Key Benefits Delivered
- Natural language access to live radiology data — worklists, patient history, reports, dashboards — without navigating multiple applications.
- 60+ MCP tools spanning clinical reporting, financial settlement, user management, scheduling, and administration.
- Role-based authorization ensuring radiologists, technologists, administrators, and typists see only permitted data.
- Multi-step query resolution — Claude automatically chains tool calls to answer complex questions.
- Persistent conversation history with MongoDB-backed session management for audit and continuity.
- 4-layer security architecture maintaining JWT-based authentication from browser through to MCP tool execution.
- Reduced navigation overhead in fast-paced radiology workflows, enabling faster clinical decision-making.
- Extensible MCP architecture allowing new tools and data sources to be added without modifying the AI engine.
Let's Get Your Project Started
Ready to discuss your project? Let's connect and explore how we can help you achieve your business goals.