Azure Image Generation MCP

Instructions

Implementation Reference

• azure-image-generation-server.js:103-151 (handler)

  The core handler function that implements the logic for the 'generate_image' tool. It destructures input arguments, selects the model, calls the appropriate image generation method (DALL-E or FLUX), handles the response, and formats the output with text and base64-encoded image data.

  async generateImage(args) {
      const startTime = Date.now();
      // Starting image generation
  
      const { prompt, model: requestedModel = 'auto', size = '1024x1024', style = 'vivid', quality = 'standard' } = args;
  
      if (!prompt) {
          throw new Error('Prompt is required for image generation');
      }
  
      try {
          const modelSelectionStart = Date.now();
          const selectedModel = this.selectModel(prompt, requestedModel);
          const modelSelectionTime = Date.now() - modelSelectionStart;
  
          // Model selection and generation setup complete
  
          const apiCallStart = Date.now();
          let result;
          if (selectedModel === 'dall-e-3') {
              result = await this.generateWithDallE(prompt, size, style, quality);
          } else {
              result = await this.generateWithFlux(prompt, size);
          }
          const apiCallTime = Date.now() - apiCallStart;
  
          const totalTime = Date.now() - startTime;
          // Image generation completed successfully
  
          return {
              content: [
                  {
                      type: 'text',
                      text: `🎨 **Your AI-Generated Image is Ready!**\n\n✨ **Created from:** "${prompt}"\n🤖 **AI Model:** ${selectedModel.toUpperCase()} ${selectedModel === 'dall-e-3' ? '(Photorealistic AI)' : '(Creative AI)'}\n📐 **Size:** ${size}\n⏱️ **Generation Time:** ${(totalTime / 1000).toFixed(1)}s\n${requestedModel === 'auto' ? `🧠 **Smart Selection:** Chose ${selectedModel.toUpperCase()} based on your prompt\n` : ''}\n🖼️ Your custom image has been generated and is displayed below. Feel free to save, share, or use it however you'd like!`
                  },
                  {
                      type: 'image',
                      mimeType: 'image/png',
                      data: result.imageData
                  }
              ]
          };
  
      } catch (error) {
          const totalTime = Date.now() - startTime;
          console.error(`❌ [MCP] Error after ${totalTime}ms:`, error);
          throw new Error(`Failed to generate image: ${error.message}`);
      }
  }

• azure-image-generation-server.js:40-72 (schema)

  The input schema for the 'generate_image' tool, defining the structure and validation for parameters: prompt (required), model (with enum and default), size, style, and quality.

  inputSchema: {
      type: 'object',
      properties: {
          prompt: {
              type: 'string',
              description: 'Describe the image you want to create in natural language. Be detailed for best results. Examples: "A serene mountain landscape at sunset", "Modern minimalist logo design", "Cute cartoon mascot for a coffee shop"'
          },
          model: {
              type: 'string',
              description: 'Choose AI model: "dall-e-3" (photorealistic, artistic), "flux" (creative, flexible), or "auto" (smart selection based on prompt)',
              enum: ['dall-e-3', 'flux', 'auto'],
              default: 'auto'
          },
          size: {
              type: 'string',
              description: 'Image dimensions: Square (1024x1024) for social media, Wide (1792x1024) for banners, Tall (1024x1792) for posters',
              enum: ['1024x1024', '1792x1024', '1024x1792'],
              default: '1024x1024'
          },
          style: {
              type: 'string',
              description: 'Visual style (DALL-E only): "vivid" for dramatic/artistic, "natural" for realistic/subdued',
              enum: ['vivid', 'natural'],
              default: 'vivid'
          },
          quality: {
              type: 'string',
              description: 'Image quality (DALL-E only): "standard" for faster generation, "hd" for higher detail',
              enum: ['standard', 'hd'],
              default: 'standard'
          }
      },
      required: ['prompt']

• azure-image-generation-server.js:35-76 (registration)

  Tool registration in the ListToolsRequestHandler. Responds with the 'generate_image' tool's name, description, and input schema when clients query available tools.

  this.server.setRequestHandler(ListToolsRequestSchema, async () => ({
      tools: [
          {
              name: 'generate_image',
              description: '🎨 Create stunning AI-generated images using Azure DALL-E 3 or FLUX models with intelligent model selection',
              inputSchema: {
                  type: 'object',
                  properties: {
                      prompt: {
                          type: 'string',
                          description: 'Describe the image you want to create in natural language. Be detailed for best results. Examples: "A serene mountain landscape at sunset", "Modern minimalist logo design", "Cute cartoon mascot for a coffee shop"'
                      },
                      model: {
                          type: 'string',
                          description: 'Choose AI model: "dall-e-3" (photorealistic, artistic), "flux" (creative, flexible), or "auto" (smart selection based on prompt)',
                          enum: ['dall-e-3', 'flux', 'auto'],
                          default: 'auto'
                      },
                      size: {
                          type: 'string',
                          description: 'Image dimensions: Square (1024x1024) for social media, Wide (1792x1024) for banners, Tall (1024x1792) for posters',
                          enum: ['1024x1024', '1792x1024', '1024x1792'],
                          default: '1024x1024'
                      },
                      style: {
                          type: 'string',
                          description: 'Visual style (DALL-E only): "vivid" for dramatic/artistic, "natural" for realistic/subdued',
                          enum: ['vivid', 'natural'],
                          default: 'vivid'
                      },
                      quality: {
                          type: 'string',
                          description: 'Image quality (DALL-E only): "standard" for faster generation, "hd" for higher detail',
                          enum: ['standard', 'hd'],
                          default: 'standard'
                      }
                  },
                  required: ['prompt']
              }
          }
      ]
  }));

• azure-image-generation-server.js:79-87 (registration)

  Tool call dispatch registration in the CallToolRequestHandler. Routes 'generate_image' tool invocations to the generateImage handler method.

  this.server.setRequestHandler(CallToolRequestSchema, async (request) => {
      const { name, arguments: args } = request.params;
  
      if (name === 'generate_image') {
          return await this.generateImage(args);
      }
  
      throw new Error(`Unknown tool: ${name}`);
  });

• azure-image-generation-server.js:91-101 (helper)

  Helper method for intelligent model selection between 'dall-e-3' and 'flux' based on explicit user request or prompt keywords.

  selectModel(prompt, requestedModel) {
      if (requestedModel !== 'auto') {
          return requestedModel;
      }
  
      // Check if user explicitly mentions DALL-E
      const lowerPrompt = prompt.toLowerCase();
      const isDalleRequest = ['dall-e', 'dalle', 'dall e'].some(keyword => lowerPrompt.includes(keyword));
  
      return isDalleRequest ? 'dall-e-3' : 'flux';
  }