power-platform-mcp-connector-suite

SKILL.md

Power Platform MCP Connector Suite

Generate comprehensive Power Platform custom connector implementations with Model Context Protocol integration for Microsoft Copilot Studio.

MCP Capabilities in Copilot Studio

Currently Supported:

✅ Tools: Functions that the LLM can call (with user approval)

✅ Resources: File-like data that agents can read (must be tool outputs)

Not Yet Supported:

❌ Prompts: Pre-written templates (prepare for future support)

Connector Generation

Create complete Power Platform connector with:
Core Files:

apiDefinition.swagger.json with x-ms-agentic-protocol: mcp-streamable-1.0

apiProperties.json with connector metadata and authentication

script.csx with custom C# transformations for MCP JSON-RPC handling

readme.md with connector documentation

MCP Integration:

POST /mcp endpoint for JSON-RPC 2.0 communication

McpResponse and McpErrorResponse schema definitions

Copilot Studio constraint compliance (no reference types, single types)

Resource integration as tool outputs (Resources and Tools supported; Prompts not yet supported)

Schema Validation & Troubleshooting

Validate schemas for Copilot Studio compliance:

✅ No reference types ($ref) in tool inputs/outputs

✅ Single type values only (not ["string", "number"])

✅ Primitive types: string, number, integer, boolean, array, object

✅ Resources as tool outputs, not separate entities

✅ Full URIs for all endpoints

Common issues and fixes:

Tools filtered → Remove reference types, use primitives

Type errors → Single types with validation logic

Resources unavailable → Include in tool outputs

Connection failures → Verify x-ms-agentic-protocol header

Context Variables

Connector Name: [Display name for the connector]

Server Purpose: [What the MCP server should accomplish]

Tools Needed: [List of MCP tools to implement]

Resources: [Types of resources to provide]

Authentication: [none, api-key, oauth2, basic]

Host Environment: [Azure Function, Express.js, etc.]

Target APIs: [External APIs to integrate with]

Generation Modes

Mode 1: Complete New Connector

Generate all files for a new Power Platform MCP connector from scratch, including CLI validation setup.

Mode 2: Schema Validation

Analyze and fix existing schemas for Copilot Studio compliance using paconn and validation tools.

Mode 3: Integration Troubleshooting

Diagnose and resolve MCP integration issues with Copilot Studio using CLI debugging tools.

Mode 4: Hybrid Connector

Add MCP capabilities to existing Power Platform connector with proper validation workflows.

Mode 5: Certification Preparation

Prepare connector for Microsoft certification submission with complete metadata and validation compliance.

Mode 6: OAuth Security Hardening

Implement OAuth 2.0 authentication enhanced with MCP security best practices and advanced token validation.

Expected Output

1. apiDefinition.swagger.json

Swagger 2.0 format with Microsoft extensions

MCP endpoint: POST /mcp with proper protocol header

Compliant schema definitions (primitive types only)

McpResponse/McpErrorResponse definitions

2. apiProperties.json

Connector metadata and branding (iconBrandColor required)

Authentication configuration

Policy templates for MCP transformations

3. script.csx

JSON-RPC 2.0 message handling

Request/response transformations

MCP protocol compliance logic

Error handling and validation

4. Implementation guidance

Tool registration and execution patterns

Resource management strategies

Copilot Studio integration steps

Testing and validation procedures

Validation Checklist

Technical Compliance

 x-ms-agentic-protocol: mcp-streamable-1.0 in MCP endpoint

 No reference types in any schema definitions

 All type fields are single types (not arrays)

 Resources included as tool outputs

 JSON-RPC 2.0 compliance in script.csx

 Full URI endpoints throughout

 Clear descriptions for Copilot Studio agents

 Authentication properly configured

 Policy templates for MCP transformations

 Generative Orchestration compatibility

CLI Validation

 paconn validate: paconn validate --api-def apiDefinition.swagger.json passes without errors

 pac CLI ready: Connector can be created/updated with pac connector create/update

 Script validation: script.csx passes automatic validation during pac CLI upload

 Package validation: ConnectorPackageValidator.ps1 runs successfully

OAuth and Security Requirements

 OAuth 2.0 Enhanced: Standard OAuth 2.0 with MCP security best practices implementation

 Token Validation: Implement token audience validation to prevent passthrough attacks

 Custom Security Logic: Enhanced validation in script.csx for MCP compliance

 State Parameter Protection: Secure state parameters for CSRF prevention

 HTTPS Enforcement: All production endpoints use HTTPS only

 MCP Security Practices: Implement confused deputy attack prevention within OAuth 2.0

Certification Requirements

 Complete metadata: settings.json with product and service information

 Icon compliance: PNG format, 230x230 or 500x500 dimensions

 Documentation: Certification-ready readme with comprehensive examples

 Security compliance: OAuth 2.0 enhanced with MCP security practices, privacy policy

 Authentication flow: OAuth 2.0 with custom security validation properly configured

Example Usage

Mode: Complete New Connector
Connector Name: Customer Analytics MCP
Server Purpose: Customer data analysis and insights
Tools Needed:
  - searchCustomers: Find customers by criteria
  - getCustomerProfile: Retrieve detailed customer data
  - analyzeCustomerTrends: Generate trend analysis
Resources:
  - Customer profiles (JSON data)
  - Analysis reports (structured data)
Authentication: oauth2
Host Environment: Azure Function
Target APIs: CRM REST API