Dataverse MCP Chatbot

Dataverse MCP Chatbot

A lightweight HTTP wrapper around a Model Context Protocol (MCP) server for Microsoft Power Platform/Dataverse. This project exposes a set of tools via a /invoke REST endpoint and includes a simple HTML chatbot UI that accepts natural-language queries and dynamically routes them to the appropriate MCP tool.

Features

• MCP Tools: Supports 8 core MCP operations:

  1. use-powerplatform-prompt — Entity overview, attribute details, query templates, relationship maps
  2. get-entity-metadata — Fetch metadata for an entity
  3. get-entity-attributes — List all attributes of an entity
  4. get-entity-attribute — Retrieve a specific attribute definition
  5. get-entity-relationships — One-to-many and many-to-many relationships
  6. get-global-option-set — Global option set definitions
  7. get-record — Fetch a record by GUID
  8. query-records — OData-filtered record queries

• Natural-Language Front End: HTML + JavaScript UI that parses free-text questions and invokes the correct MCP tool.

• Environment Configuration: Uses environment variables for all Power Platform credentials.

• Express + CORS: Simple HTTP server for integration with any front-end.

Prerequisites

• Node.js >= 16.x

• A Dataverse / Power Platform environment and service principal (app registration) with:

  • Tenant ID
  • Client (Application) ID
  • Client Secret
  • Organization URL (e.g. https://<yourorg>.crm.dynamics.com)

Installation

1. Clone the repository

   git clone https://github.com/your-org/powerplatform-mcp-chatbot.git
   cd powerplatform-mcp-chatbot
   

2. Install dependencies

   npm install
   

3. Build the TypeScript

   npm run build
   

Configuration

Create a .env file in the project root (or set environment variables in your shell) with:

POWERPLATFORM_URL=https://yourorg.crm.dynamics.com
POWERPLATFORM_CLIENT_ID=your-client-id
POWERPLATFORM_CLIENT_SECRET=your-client-secret
POWERPLATFORM_TENANT_ID=your-tenant-id
PORT=3000           # Optional: HTTP server port

Note: Ensure these values are kept secure and not committed to version control.

Running the Server

After building, start the HTTP wrapper:

node dist/server.js

You should see:

✅ PowerPlatform MCP HTTP server running at http://localhost:3000/invoke

REST Invoke Endpoint

All MCP tools are exposed via a single POST endpoint:

POST http://localhost:3000/invoke
Content-Type: application/json

{
  "tool": "tool-name",
  "parameters": { /* tool-specific parameters */ }
}

Sample Requests

1. Entity Overview

   curl -X POST http://localhost:3000/invoke \
     -H 'Content-Type: application/json' \
     -d '{"tool":"use-powerplatform-prompt","parameters":{"promptType":"ENTITY_OVERVIEW","entityName":"account"}}'
   

2. Get Entity Metadata

   curl -X POST http://localhost:3000/invoke \
     -d '{"tool":"get-entity-metadata","parameters":{"entityName":"contact"}}'
   

3. Query Records

   curl -X POST http://localhost:3000/invoke \
     -d '{"tool":"query-records","parameters":{"entityNamePlural":"contacts","filter":"statecode eq 0","maxRecords":10}}'
   

Refer to the full tool list above for parameter details.

HTML Chatbot UI

A demo front-end is provided in chatbot.html. It listens for natural-language queries, maps them to the correct tool, and displays results inline.

1. Serve chatbot.html (e.g., open directly in Chrome or host via any static server).

2. Ask questions such as:

   • “Tell me about the Account entity”
   • “List fields in Lead”
   • “Details of field revenue on Account”
   • “Find active contacts where statecode eq 0”

Troubleshooting

• CORS Errors: Ensure you run the HTML from a local or hosted file with CORS enabled (the server uses cors() by default).
• Authentication: Verify your service principal has Dataverse API permissions and the environment URL is correct.
• Errors: Check console output from server.js for stack traces and error messages.