Monitor credit consumption with AI Builder Credits MCP connector

March 20, 2026

With Microsoft’s announcement of the end of AI Builder credits, monitoring your credit consumption has become critical. Seeded credits from premium licenses end November 1, 2026, and add-on credits will stop being available for renewal. When credits run out, AI Builder features fall back to Copilot Studio Credits—and if those are exhausted too, your automations stop working.

A customer recently asked me how they could proactively track their AI Builder usage before hitting these limits. The answer: give Copilot Studio agents direct access to the consumption data. This MCP connector queries the msdyn_aievents table in Dataverse so agents can report on credit usage, identify high-consumption models, and help plan the transition to Copilot Credits.

What this connector does

The AI Builder Credits connector queries the msdyn_aievents table in Dataverse to retrieve credit consumption data. It provides five MCP tools that let agents:

• List recent AI events with credit details
• Get credit summaries grouped by model and source
• Track daily usage trends
• Retrieve specific event details
• List AI Builder models in the environment

Available tools

Prerequisites

Before you can use this connector, you need:

• A Power Platform environment with AI Builder usage
• An Azure AD app registration with delegated permissions for Dataverse user access
• Admin consent granted for the permissions

Setting up the connector

Register an Azure AD app

1. Go to the Azure portal > Microsoft Entra ID > App registrations
2. Register a new application or select an existing one
3. Under Authentication, add https://global.consent.azure-apim.net/redirect as a redirect URI
4. Copy the Application (client) ID into the connector’s apiProperties.json clientId field

Import the connector

1. Download the connector files from the GitHub repository
2. In the Power Platform admin center, go to Custom connectors
3. Import the connector using the swagger and properties files
4. Update the clientId in the properties file with your app registration ID

Tool details

list_ai_events

List recent AI Builder events with credit consumption details.

Parameters:

• top (integer) - Maximum events to return (default: 25, max: 100)
• source (string) - Filter by source: PowerAutomate, PowerApps, API, or CopilotStudio
• fromDate (string) - Filter events from this date (ISO 8601)

Returns:

• Event ID, model name, credits consumed, date, source, status, automation name

get_credit_summary

Get a summary of credit consumption grouped by model and source.

Parameters:

• fromDate (string) - Start date (defaults to first of current month)
• toDate (string) - End date (defaults to today)

Returns:

• Total events and credits
• Breakdown by AI model
• Breakdown by source (Power Automate, Power Apps, API, Copilot Studio)

get_daily_usage

Get credit consumption grouped by day for trend analysis.

Parameters:

• days (integer) - Days to look back (default: 7, max: 30)

Returns:

• Daily usage with event count and credits per day

get_ai_event

Get details of a specific AI Builder event.

Parameters:

• eventId (string, required) - The AI event GUID

Returns:

• Full event details including model template, automation link, data info

list_ai_models

List AI Builder models in the environment.

Parameters:

• top (integer) - Maximum models to return (default: 50)

Returns:

• Model ID, name, template type, status, creation date

The data source

The connector queries the Dataverse msdyn_aievents table, which stores:

• msdyn_creditconsumed - Credits used per action
• msdyn_AIModelId - Reference to the AI model
• msdyn_processingdate - When the action occurred
• msdyn_consumptionsource - 0=Power Automate, 1=Power Apps, 2=API, 3=Copilot Studio
• msdyn_processingstatus - 0=Processed
• msdyn_automationname - Name of the flow or app that triggered the action

How the MCP implementation works

The connector uses a custom script connector pattern with the MCP protocol. The swagger definition includes the x-ms-agentic-protocol: mcp-streamable-1.0 annotation to enable MCP support:

{
  "paths": {
    "/": {
      "post": {
        "operationId": "InvokeMCP",
        "x-ms-agentic-protocol": "mcp-streamable-1.0",
        "parameters": [
          {
            "name": "body",
            "in": "body",
            "required": true,
            "schema": {
              "type": "object"
            }
          }
        ]
      }
    }
  }
}

The script.csx file handles all MCP protocol operations and routes tool calls to the appropriate handlers. Each tool queries Dataverse using OData and returns structured JSON responses.

Known limitations

• Environment-scoped data - You need separate connections to monitor multiple environments
• Periodic computation - Credit consumption data is computed periodically (typically daily) by the platform
• Data retention - Historical data retention depends on your Dataverse capacity settings

Use cases

This connector is useful for:

• Cost management - Track which models and automations consume the most credits
• Capacity planning - Analyze usage trends to predict future credit needs
• Troubleshooting - Identify which automations are causing unexpected credit consumption
• Reporting - Generate usage reports for stakeholders

Resources

• Full source code on GitHub
• AI Builder Activity Monitoring
• AI Event (msdyn_AIEvent) table reference
• AI Builder Credit Management