Copilot Studio Analytics connector for Power Platform

February 09, 2026

Copilot Studio captures rich analytics data in Dataverse, but getting that data into flows, apps, or other agents isn’t straightforward. This custom connector wraps those Dataverse tables with purpose-built operations for conversation analytics, topic performance, session outcomes, customer satisfaction scores, and full transcript parsing. It also exposes those same analytics as MCP tools so a Copilot Studio agent can report on its own performance conversationally.

You can find the complete code in my SharingIsCaring repository.

What it does

The connector queries four Dataverse tables—bot, conversationtranscript, botcomponent, and msdyn_botsession—and computes analytics server-side in the connector script. Instead of building complex Dataverse queries and post-processing in your flow, call a single operation and get structured metrics back.

Operations

Bot management

Transcripts

Transcript parsing converts the raw JSON content into structured data including user/bot message counts, topics triggered, escalation status, and individual messages with roles and timestamps.

Analytics

Session and conversation data

Analytics deep dive

Bot conversation analytics

Call GetBotConversationAnalytics with a bot ID and optional date range to get:

• Total conversations and daily breakdown
• Average messages per conversation (total and user-only)
• Average duration in seconds
• Escalation rate and escalated count

{
  "botId": "12345678-...",
  "totalConversations": 150,
  "averageMessagesPerConversation": 8.5,
  "averageDurationSeconds": 245.5,
  "escalationRate": 12.5,
  "escalatedCount": 19,
  "conversationsByDay": [
    { "date": "2025-01-15", "count": 25 }
  ]
}

Topic analytics

GetTopicAnalytics returns the top N topics ranked by trigger count, each with completion and escalation rates:

{
  "topics": [
    {
      "topicName": "Order Status",
      "triggerCount": 450,
      "completionRate": 92.5,
      "escalationRate": 5.2
    }
  ]
}

Session analytics

GetSessionAnalytics breaks sessions into outcomes:

{
  "totalSessions": 500,
  "averageSessionDurationSeconds": 180.5,
  "outcomeBreakdown": [
    { "outcome": "resolved", "count": 400, "percentage": 80.0 },
    { "outcome": "escalated", "count": 75, "percentage": 15.0 },
    { "outcome": "abandoned", "count": 25, "percentage": 5.0 }
  ]
}

CSAT analytics

GetCSATAnalytics returns satisfaction scores with rating distribution and recent feedback:

{
  "totalResponses": 120,
  "averageRating": 4.2,
  "csatScore": 85.0,
  "ratingDistribution": [
    { "rating": 5, "count": 55, "percentage": 45.8 },
    { "rating": 4, "count": 42, "percentage": 35.0 }
  ],
  "recentFeedback": [
    { "transcriptId": "abc123...", "rating": 5, "comment": "Very helpful!" }
  ]
}

MCP protocol

The connector includes a /mcp endpoint with seven JSON-RPC 2.0 tools for Copilot Studio agent integration:

This means a Copilot Studio agent can answer questions about its own analytics:

User: “How many conversations did the support bot handle last week?” Agent: Calls get_conversation_analytics with date range and returns the count

User: “Which topics have the highest escalation rate?” Agent: Calls get_topic_analytics and highlights topics with high escalation

User: “What’s our CSAT score for January?” Agent: Calls get_csat_analytics with the date range

Setup

1. Azure AD app registration

1. Create or reuse an app registration in the Azure Portal.
2. Add API permission: Dynamics CRM → user_impersonation.
3. Grant admin consent.
4. Note the Application (client) ID and Directory (tenant) ID.

2. Import the connector

1. Import apiDefinition.swagger.json as a custom connector in Power Platform.
2. Enable custom code and upload script.csx.
3. Create a connection with your tenant and Dataverse domain.

3. Dataverse security

Ensure your user has System Administrator, System Customizer, or a custom role with read access to the bot and conversationtranscript tables.

Power Automate examples

Daily conversation count

1. Call ListConversationTranscripts with filter: createdon ge @{formatDateTime(utcNow(), 'yyyy-MM-dd')}T00:00:00Z
2. Use the Length expression on the value array.

Export transcripts to SharePoint

1. ListConversationTranscripts for a specific bot.
2. Apply to each on value.
3. Create file in SharePoint with transcript content.

Analytics summary email

1. Call GetConversationAnalytics with a date filter.
2. Compose an HTML email body with the statistics.
3. Send the email.

Common OData filters

Filter by bot:

_bot_value eq '12345678-1234-1234-1234-123456789012'

Filter by date range:

createdon ge 2024-01-01T00:00:00Z and createdon le 2024-01-31T23:59:59Z

Application Insights (optional)

Update the connection string in script.csx to enable telemetry:

private static readonly string AppInsightsConnectionString = "InstrumentationKey=your-key-here;...";

Events logged: CopilotStudioAnalytics_Request, CopilotStudioAnalytics_Response, CopilotStudioAnalytics_Error.

Try it yourself

The complete connector code is available in my SharingIsCaring repository:

• apiDefinition.swagger.json — OpenAPI specification
• script.csx — Connector script with analytics and MCP
• apiProperties.json — Connector metadata
• readme.md — Full documentation

Resources

• Copilot Studio documentation
• Dataverse Web API reference
• OData query options
• Model Context Protocol specification
• Power Platform Custom Connectors
• Application Insights telemetry for connectors