Microsoft Foundry MAI Image MCP connector for Power Platform

April 06, 2026

MAI-Image-2 launched April 2, 2026, as part of Microsoft’s three-model MAI announcement alongside MAI-Transcribe-1 and MAI-Voice-1. It ranks in the top 3 on the Arena.ai text-to-image leaderboard and delivers at least 2x faster generation times on Foundry and Copilot based on real-world production traffic data. Accurate skin tones, natural lighting, and clear in-image text rendering—built for photographers, designers, and visual storytellers who need campaign-ready imagery from text prompts.

This connector brings MAI-Image-2 and gpt-image-1.5 into Power Platform with two MCP tools for Copilot Studio and two REST operations for Power Automate and Power Apps. A borrowed chat completion operation lets agents describe, interpret, or discuss generated images within a conversation.

Full source: GitHub repository

The models

Both models accept the same parameters—prompt, size, quality, style, response format, and output format—through a single operation. Select the model at request time.

Tools

MCP tools for Copilot Studio

How it works

User: "Create a product photo of a coffee mug on a wooden table
       with morning sunlight streaming through a window"

1. Orchestrator calls generate_image({
     prompt: "Product photo of a white ceramic coffee mug on a
              rustic wooden table, warm morning sunlight streaming
              through a curtained window, shallow depth of field,
              photorealistic",
     quality: "high",
     style: "natural"
   })

   → Returns:
     url: "https://..."  (temporary, expires in 60 minutes)
     revised_prompt: "A white ceramic coffee mug sitting on..."
     image_count: 1
     usage: { input_tokens: 42, output_tokens: 1024 }

User: "Describe this image and suggest how to improve the prompt"

2. Orchestrator calls chat_completion({
     prompt: "Describe this product photo and suggest prompt
              improvements for better composition",
     system_prompt: "You are an expert product photographer"
   })

   → Returns suggestions for prompt refinement

REST operations for Power Automate and Power Apps

Parameter reference

Size options

Generate Photorealistic Image response

Use cases

Marketing content and campaign imagery: Generate on-brand product photos, social media graphics, and ad visuals directly from Power Automate flows. Pair with the chat completion tool to iterate on prompts until the output matches creative direction.

Product visualization and mockups: Create realistic product renders from text descriptions before investing in physical photography. Adjust lighting, backgrounds, and compositions through prompt parameters.

Branded asset creation: Generate social media posts, banner ads, and presentation graphics with consistent visual style. Use the natural style for photorealistic results or vivid for bold, attention-grabbing imagery.

Presentation graphics with in-image text: MAI-Image-2 handles text rendering better than most image generation models. Generate slides, infographics, and labeled diagrams with readable text baked into the image.

Creative concept exploration: Use a Copilot Studio agent to brainstorm visual concepts interactively. Generate variations, get the agent to describe what it sees via chat_completion, and refine prompts in a conversational loop.

Prerequisites

1. An Azure subscription with access to Microsoft Foundry
2. Deploy an image generation model from the Foundry Model Catalog:
   • MAI-Image-2 — for photorealistic image generation
   • gpt-image-1.5 — for general image generation
3. Note the Resource Name (for example, my-foundry-resource from https://my-foundry-resource.services.ai.azure.com) and API Key from the deployment

Setting up the connector

1. Deploy an image model in Microsoft Foundry

1. Go to the Foundry Model Catalog
2. Search for MAI-Image-2 or gpt-image-1.5 and deploy the model
3. Copy the Resource Name and API Key from the deployment page

2. Create the custom connector

1. Go to Power Platform Maker Portal
2. Navigate to Custom connectors > + New custom connector > Import an OpenAPI file
3. Upload apiDefinition.swagger.json
4. On the Security tab:
   • Authentication type: API Key
   • Parameter label: API Key
   • Parameter name: api-key
   • Parameter location: Header
5. On the Code tab:
   • Enable Code
   • Upload script.csx
6. Select Create connector

3. Create a connection

1. Select Test > + New connection
2. Enter your Resource Name and API Key
3. Select Create connection

4. Test the connector

Test GeneratePhotorealisticImage with a simple prompt:

{
  "prompt": "A golden retriever sitting in a field of wildflowers at sunset, photorealistic"
}

The response includes an image URL you can open in a browser to verify the output.

5. Add to Copilot Studio

1. In Copilot Studio, open your agent
2. Add this connector as an action—Copilot Studio detects the MCP endpoint via x-ms-agentic-protocol
3. Test with prompts like “Generate a product photo of…” or “Create a banner image for…”

Known limitations

• Image URLs expire after 60 minutes—use b64_json response format for persistent storage
• Base64 responses can be large (1 MB or more for high-quality PNG)
• Maximum 10 images per request
• Maximum 32,000 character prompt
• The chat_completion tool is borrowed from the parent Microsoft Foundry connector and requires a chat-capable model deployed to the same resource

Files

Resources

• MAI Image connector source code
• Microsoft Foundry Model Catalog
• Microsoft Foundry API
• MAI-Image-2 model card (PDF)
• 3 new MAI models announcement
• MAI Playground (US only)
• Arena.ai text-to-image leaderboard