PI API MCP Server

A Model Context Protocol (MCP) server that provides standardized tools and resources for interacting with the PI Dashboard API. This implementation enables Claude and other MCP-compatible AI assistants to securely access and manage PI Dashboard resources, including categories and charts.

Utilizing PI with MCP

The following demonstrates typical usage scenarios for this MCP Server after setup completion.

Initial Authentication:

• Execute the following instructions to establish a connection:

Chart Analysis:

• If chart ID 450 contains metadata information, use the following prompt:

Example Output:

Installation

Installing via Smithery

To install pi-api-mcp-server for Claude Desktop automatically via Smithery:

Installation - Using Docker (Recommended)

• No MCP Server configuration needed
• MCP client configuration file setup:

Important Note: If the --api-url parameter is not provided at initialization, the server will prompt you to configure the API URL using the set-api-url tool before executing any operations. This design enables flexible configuration in environments where the URL is not predetermined at startup.

Configuration File Location

Access your Claude for Desktop application configuration at:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: Use other tools for now. e.g. Cline - ask it to show you the MCP config file

Available Tools

Schema Discovery

• get-filterable-attributes: Get the list of attributes that can be used for filtering by examining a sample entity
  Get the filterable attributes for chart entities

Connection Management

• check-connection: Check if the current API URL and authentication are valid
• set-api-url: Configure the base API URL for all requests
  Set the API URL to http://localhost:8224/pi/api/v2

Authentication

• authenticate: Get guidance on authentication options
• authenticate-with-credentials: Authenticate with username and password (last resort option)
• keep-session-alive: Verify and refresh the current authentication token (also used for token-based authentication)
• logout: Invalidate the current token and end the session
• set-organization: Set the organization ID for subsequent requests

Categories

• list-categories: List all categories with filtering support
• get-category: Get a category by ID
• create-category: Create a new category
• update-category: Update an existing category
• delete-category: Delete a category
• list-category-objects: List all objects for a specific category

Charts

• list-charts: List all charts with filtering support
• get-chart: Get a chart by ID
• delete-chart: Delete a chart
• export-chart: Export a chart in various formats

Available Resources

• auth://status: Get authentication status
• categories://list: List all categories
• categories://{id}: Get a specific category
• categories://{categoryId}/objects: Get objects for a specific category
• charts://list: List all charts
• charts://{id}: Get a specific chart
• charts://{id}/export/{format}: Export a chart in a specific format

Available Prompts

• analyze-categories: Analyze categories in the dashboard
• analyze-charts: Analyze charts in the dashboard
• compare-charts: Compare data between two charts
• category-usage-analysis: Analyze how categories are being used in charts
• use-filters: Shows how to use filters effectively with this API

Claude Integration Examples

Here are some example queries to use with Claude after connecting the server:

Set the API URL

Authentication

Working with Categories

Working with Charts

Using Filters

Using Analysis Prompts

Development

Local Execution

• Note: you can make use of start.sh to run the dev server as well.

NPM Installation

MCP Client Configuration

Integration with Claude for Desktop:

Node.js Implementation

• Execute the instructions in the "Local Execution" section
• Ensure ./dependencies.sh has been executed to install required dependencies
• Implement the following configuration (Note: "@mingzilla/pi-api-mcp-server" references the package installed through "Local Execution")

Local Development

• run the server using ./start.sh
• set the config with the path to the build/index.js file

Development Check List

• update code -> start local server -> test local server with file path to index.js
• update readme.md file -> change the mcpServers config section: docker + node + npx
• ./publish.sh - publish to npm
• ./dockerBuild.sh -> ./dockerPublish.sh (edit version number to match package.json) -> test docker config
• push code to github

License

MIT License

Author

Ming Huang (mingzilla)