Cover image
Try Now
2025-03-18

PowerPlatform Model Context Protocol server

3 years

Works with Finder

1

Github Watches

2

Github Forks

8

Github Stars

PowerPlatform MCP Server

A Model Context Protocol (MCP) server that provides intelligent access to PowerPlatform/Dataverse entities and records. This tool offers context-aware assistance, entity exploration and metadata access.

Key features:

  • Rich entity metadata exploration with formatted, context-aware prompts
  • Advanced OData query support with intelligent filtering
  • Comprehensive relationship mapping and visualization
  • AI-assisted query building and data modeling through AI agent
  • Full access to entity attributes, relationships, and global option sets

Installation

You can install and run this tool in two ways:

Option 1: Install globally

npm install -g powerplatform-mcp

Then run it:

powerplatform-mcp

Option 2: Run directly with npx

Run without installing:

npx powerplatform-mcp

Configuration

Before running, set the following environment variables:

# PowerPlatform/Dataverse connection details
POWERPLATFORM_URL=https://yourenvironment.crm.dynamics.com
POWERPLATFORM_CLIENT_ID=your-azure-app-client-id
POWERPLATFORM_CLIENT_SECRET=your-azure-app-client-secret
POWERPLATFORM_TENANT_ID=your-azure-tenant-id

Usage

This is an MCP server designed to work with MCP-compatible clients like Cursor, Claude App and GitHub Copilot. Once running, it will expose tools for retrieving PowerPlatform entity metadata and records.

Available Tools

  • get-entity-metadata: Get metadata about a PowerPlatform entity
  • get-entity-attributes: Get attributes/fields of a PowerPlatform entity
  • get-entity-attribute: Get a specific attribute/field of a PowerPlatform entity
  • get-entity-relationships: Get relationships for a PowerPlatform entity
  • get-global-option-set: Get a global option set definition
  • get-record: Get a specific record by entity name and ID
  • query-records: Query records using an OData filter expression
  • use-powerplatform-prompt: Use pre-defined prompt templates for PowerPlatform entities

MCP Prompts

The server includes a prompts feature that provides formatted, context-rich information about PowerPlatform entities.

Available Prompt Types

The use-powerplatform-prompt tool supports the following prompt types:

  1. ENTITY_OVERVIEW: Comprehensive overview of an entity
  2. ATTRIBUTE_DETAILS: Detailed information about a specific entity attribute
  3. QUERY_TEMPLATE: OData query template for an entity with example filters
  4. RELATIONSHIP_MAP: Visual map of entity relationships

Examples

Entity Overview Prompt

// Example client code
await mcpClient.invoke("use-powerplatform-prompt", {
  promptType: "ENTITY_OVERVIEW",
  entityName: "account"
});

Output:

## Power Platform Entity: account

This is an overview of the 'account' entity in Microsoft Power Platform/Dataverse:

### Entity Details
- Display Name: Account
- Schema Name: Account
- Description: Business that represents a customer or potential customer
- Primary Key: accountid
- Primary Name: name

### Key Attributes
- name: String (Account Name)
- emailaddress1: String (Email)
- telephone1: String (Main Phone)
- address1_city: String (City)
- statecode: Status (Status)
- accountnumber: String (Account Number)
- industrycode: OptionSetValue (Industry)
- revenue: Money (Annual Revenue)
- ownerid: Owner (Owner)
- createdon: DateTime (Created On)

### Relationships
- One-to-Many Relationships: 42
- Many-to-Many Relationships: 8

You can query this entity using OData filters against the plural name.

Attribute Details Prompt

await mcpClient.invoke("use-powerplatform-prompt", {
  promptType: "ATTRIBUTE_DETAILS",
  entityName: "account",
  attributeName: "revenue"
});

Output:

## Attribute: revenue

Details for the 'revenue' attribute of the 'account' entity:

- Display Name: Annual Revenue
- Description: Annual revenue for the account
- Type: Money
- Format: Currency
- Is Required: No
- Is Searchable: true

### Usage Notes
- Data Type: Money
- Required: No
- Max Length: N/A

Query Template Prompt

await mcpClient.invoke("use-powerplatform-prompt", {
  promptType: "QUERY_TEMPLATE",
  entityName: "account"
});

Output:

## OData Query Template for accounts

Use this template to build queries against the accounts entity:
accounts?$select=name,emailaddress1,telephone1, address1_city,statecode&$filter=name eq 'Example'&$orderby=name asc&$top=50

Common Filter Examples

  • Equals: name eq 'Contoso'
  • Contains: contains(name, 'Contoso')
  • Greater than date: createdon gt 2023-01-01T00:00:00Z
  • Multiple conditions: name eq 'Contoso' and statecode eq 0

#### Relationship Map Prompt

```javascript
await mcpClient.invoke("use-powerplatform-prompt", {
  promptType: "RELATIONSHIP_MAP",
  entityName: "contact"
});

Output:

## Relationship Map for contact

This shows all relationships for the 'contact' entity:

### One-to-Many Relationships (contact as Primary)
- contact_activity_parties: contact (1) → activityparty (N)
- contact_connections1: contact (1) → connection (N)
- contact_connections2: contact (1) → connection (N)
- contact_customer_contacts: contact (1) → contact (N)
- contact_master_contact: contact (1) → contact (N)

### One-to-Many Relationships (contact as Related)
- account_primary_contact: account (1) → contact (N)
- customer_contacts: customer (1) → contact (N)
- system_user_contacts: systemuser (1) → contact (N)

### Many-to-Many Relationships
- contactleads_association: contact (N) ↔ lead (N)
- contactopportunities_association: contact (N) ↔ opportunity (N)
- contactquotes_association: contact (N) ↔ quote (N)

Prompt Examples

Here are some examples of how you can interact with assistant using the PowerPlatform MCP prompts:

1. Entity Understanding

You can ask assistant about entities in natural language:

User: "Tell me about the Account entity in our PowerPlatform environment"

Response:
[uses ENTITY_OVERVIEW prompt internally]

Here's what I found about the Account entity:
- It's a core business entity representing customers and potential customers
- The primary identifier is 'accountid' and the primary name field is 'name'
- Key fields include:
  * name (Account Name)
  * emailaddress1 (Email)
  * telephone1 (Main Phone)
  * revenue (Annual Revenue)
- It has 42 one-to-many relationships and 8 many-to-many relationships

2. Building Queries

You can ask for help constructing queries:

User: "Help me write a query to find active accounts with revenue over $1M in California"

Response:
[uses QUERY_TEMPLATE prompt internally]

Here's the OData query you need:
accounts?$select=name,revenue,statecode,address1_stateorprovince
&$filter=statecode eq 0 
    and revenue gt 1000000 
    and address1_stateorprovince eq 'California'
&$orderby=revenue desc

This query:
1. Selects relevant fields
2. Filters for active accounts (statecode eq 0)
3. Checks revenue > $1M
4. Filters for California location
5. Orders by revenue highest to lowest

3. Understanding Relationships

You can explore entity relationships naturally:

User: "How are Accounts and Contacts related in our system?"

Response:
[uses RELATIONSHIP_MAP prompt for both entities internally]

Here are the key relationships between Accounts and Contacts:

1. Primary Contact (One-to-Many):
   - account_primary_contact: Account (1) → Contact (N)
   - An account can have many contacts, but one designated as primary

2. Customer Relationship:
   - contact_customer_accounts: Contact (N) ← Account (1)
   - Contacts can be associated with customer accounts

3. Additional Relationships:
   - accountleads_association: For lead tracking
   - account_connections: For relationship management

4. Field/Attribute Investigation

You can ask about specific fields:

User: "What kind of field is the revenue field on Account?"

Response:
[uses ATTRIBUTE_DETAILS prompt internally]

The revenue field on Account is:
- Data Type: Money
- Format: Currency
- Not required
- Searchable: Yes
- Used for: Tracking annual revenue for the account
- Can be used in advanced find and filtered views

5. Complex Data Modeling

You can get help with data modeling decisions:

User: "I need to design a solution for tracking product reviews. Which entities should I relate to?"

Response:
[uses multiple ENTITY_OVERVIEW and RELATIONSHIP_MAP prompts internally]

Based on the existing data model, here's a suggested approach:

1. Product (existing entity):
   - Already has relationships with:
     * Price lists
     * Product categories
     * Inventory units

2. New Review entity should relate to:
   - Product (N:1) - Each review is for one product
   - Contact (N:1) - Reviews are written by contacts
   - Account (N:1) - Optional link to company if B2B review

Here's a sample lookup field structure...
[continues with detailed implementation guidance]

6. Debugging Assistance

You can get help troubleshooting:

User: "Why am I getting 'invalid property' when querying account.customertypecode?"

Response:
[uses ATTRIBUTE_DETAILS prompt internally]

I see the issue. The field 'customertypecode' is actually named 'customertype' in your environment. Here's the correct field information:
- Logical Name: customertype
- Type: OptionSetValue
- Valid Values:
  * 1 = Competitor
  * 2 = Consultant
  * 3 = Customer
  * 4 = Investor
  * 5 = Partner
  * 6 = Influencer

Try updating your query to use 'customertype' instead of 'customertypecode'.

These examples show how AI assistant can leverage the MCP prompts to provide context-aware, accurate assistance for PowerPlatform development tasks. The AI understands your environment's specific configuration and can help with both simple queries and complex architectural decisions.

License

MIT

相关推荐

  • NiKole Maxwell
  • I craft unique cereal names, stories, and ridiculously cute Cereal Baby images.

  • Emmet Halm
  • Converts Figma frames into front-end code for various mobile frameworks.

  • https://suefel.com
  • Latest advice and best practices for custom GPT development.

  • Khalid kalib
  • Write professional emails

  • Yasir Eryilmaz
  • AI scriptwriting assistant for short, engaging video content.

  • Daren White
  • A supportive coach for mastering all Spanish tenses.

  • momi
  • Provides initial medical assessments and advice.

  • https://maiplestudio.com
  • Find Exhibitors, Speakers and more

  • Joshua Armstrong
  • Confidential guide on numerology and astrology, based of GG33 Public information

  • J. DE HARO OLLE
  • Especialista en juegos de palabras en varios idiomas.

  • huahuayu
  • A unified API gateway for integrating multiple etherscan-like blockchain explorer APIs with Model Context Protocol (MCP) support for AI assistants.

  • deemkeen
  • control your mbot2 with a power combo: mqtt+mcp+llm

  • zhaoyunxing92
  • 本项目是一个钉钉MCP(Message Connector Protocol)服务,提供了与钉钉企业应用交互的API接口。项目基于Go语言开发,支持员工信息查询和消息发送等功能。

  • ShrimpingIt
  • Micropython I2C-based manipulation of the MCP series GPIO expander, derived from Adafruit_MCP230xx

  • justmywyw
  • Short and sweet example MCP server / client implementation for Tools, Resources and Prompts.

    Reviews

    4 (1)
    Avatar
    user_z5ILbMPu
    2025-04-15

    The Latex Mathml Server, developed by HappyAny, is an incredible solution for efficiently converting LaTeX to MathML. It's user-friendly and integrates seamlessly into any workflow, enhancing productivity and precision in rendering mathematical content. Highly recommended for educators, researchers, and students who rely on accurate mathematical representations. More details can be found on the official page.