Sign in

The Flow AI Gateway: Advanced Industrial Data Integration via Model Context Protocol

Introduction to Model Context Protocol (MCP)

The Model Context Protocol is an open standard that enables developers to build secure, two-way connections between their data sources and AI-powered tools. MCP (Model Context Protocol), a new protocol open-sourced by Anthropic, provides standardized interfaces for LLM applications to integrate with external data sources and tools.

The protocol creates a standardized architecture where AI applications can dynamically discover and interact with external tools and data sources. With MCP, Claude and other AI applications can dynamically discover and identify which tools to use for a given task.

What Makes MCP Powerful

MCP transforms how AI assistants interact with industrial systems by:

  • Standardized Communication: Providing a unified protocol for data exchange between AI models and external systems
  • Dynamic Tool Discovery: Allowing AI assistants to automatically discover available tools and data sources
  • Secure Integration: Ensuring safe, controlled access to sensitive industrial data
  • Real-time Interaction: Enabling immediate responses to data queries and system commands

Flow AI Gateway: Industrial MCP Server

The Flow AI Gateway represents a specialized MCP server designed specifically for manufacturing, utilities, and process industries. Built on the Flow Software platform, it serves as a central single-version-of-the-truth analytics framework that connects AI assistants directly to industrial data systems.

Core Architecture

The Flow AI Gateway operates as an MCP server that exposes industrial data through a comprehensive set of tools organized around the Flow data model:

 

Key Design Considerations

Performance Optimization Strategy

The Flow AI Gateway implements a hierarchical data access strategy for optimal performance:

  1. Primary Route: Use pre-calculated measure data (get_measure_calendar_data, get_measure_event_data) for business reporting
  2. Fallback Route: Use real-time tag aggregation only when measure data is unavailable
  3. Specialized Route: Use raw tag data for equipment monitoring and troubleshooting

Error Handling and Recovery

The system implements sophisticated error handling with contextual guidance:

  • Error Analysis: Automatic categorization of errors (InputError, NotFound, ValidationError, SystemError)
  • Context Retrieval: Errors include promptHint fields pointing to specific documentation
  • Recovery Actions: Step-by-step guidance for parameter correction and retry strategies

Security and Access Control

The Flow AI Gateway maintains industrial-grade security:

  • Controlled Data Access: Respects existing Flow Software permissions. Thus, if security had been applied to the folder hierarchy in your Flow model, these will be excluded from tool usage. *Adding configurable security permission is on the roadmap to be enhanced.
  • Parameter Validation: Strict input validation to prevent injection attacks

  • Malicious prompts: Please be aware that there is currently no safeguard on data retrieval. This has the potential to blow up your chat's context window if a large time range is returned by the tool.

     

Available Tool Categories

The following tool sets are made available via the Flow AI Gateway:

1. Discovery and Navigation Tools

Search Capabilities:

  • get_search_result: Searches across names, descriptions, properties, and fields
  • get_folders: Retrieves organizational hierarchy
  • get_folder_folders: Navigates child folders
  • get_folder_metrics: Finds KPIs for specific equipment
  • get_folder_events: Discovers operational events

System Exploration:

  • get_metrics: Discovers all performance indicators
  • get_events: Lists all operational activities
  • get_measures: Examines individual data points

2. Data Retrieval Tools

Primary Data Access (Recommended):

  • get_measure_calendar_data: Pre-calculated hourly, daily, weekly totals and aggregations
  • get_measure_event_data: Business metrics contextualized by production runs and batches

Fallback Data Access:

  • get_tag_calendar_aggregated_data: Real-time aggregation from raw sensor data
  • get_tag_event_aggregated_data: Custom event period analysis
  • get_tag_raw_data: Individual sensor readings and state changes

3. Calendar and Time Management

Calendar Configuration:

  • get_calendars: Retrieves working patterns and shift definitions
  • get_calendar_shifts: Detailed shift schedules and rotations
  • get_calendar_start: Human-readable calendar configuration
  • get_calendar_shift_interval_dates: Converts shift intervals to time ranges

4. Model Structure Tools

Relationship Discovery:

  • get_metric_measures: Components that make up KPIs
  • get_event_measures: Data points associated with operational events

5. Context and Documentation

System Guidance:

  • get_context_for_key: Retrieves detailed documentation and troubleshooting guidance
  • get_execution_plan: Generates optimized workflows for complex queries