Amazon Bedrock MCP Server

Instructions

Implementation Reference

• src/index.ts:182-311 (handler)

  Handler for the generate_image tool. Validates input using GenerateImageSchema, invokes AWS Bedrock Nova model via InvokeModelCommand, processes the response to extract base64 images, and returns MCP-formatted content with images.

  server.setRequestHandler(CallToolRequestSchema, async (request) => {
    if (request.params.name !== "generate_image") {
      throw new McpError(
        ErrorCode.MethodNotFound,
        `Unknown tool: ${request.params.name}`
      );
    }
  
    try {
      // Validate and parse input
      const args = GenerateImageSchema.parse(request.params.arguments);
      
      server.sendLoggingMessage({
        level: "info",
        data: `Configuration: ${JSON.stringify({
          width: args.width,
          height: args.height,
          quality: args.quality,
          numberOfImages: args.numberOfImages,
          cfgScale: args.cfg_scale,
          seed: args.seed
        })}`,
      });
  
      const progressToken = request.params._meta?.progressToken;
  
      server.sendLoggingMessage({
        level: "info",
        data: "Sending request to Bedrock API...",
      });
  
      const response = await bedrock.send(new InvokeModelCommand({
        modelId: NOVA_MODEL_ID,
        contentType: "application/json",
        accept: "application/json",
        body: JSON.stringify({
          taskType: GENERATION_TYPES.TEXT_TO_IMAGE,
          textToImageParams: {
            text: args.prompt,
            negativeText: args.negativePrompt || undefined,
          },
          imageGenerationConfig: {
            numberOfImages: args.numberOfImages,
            height: args.height,
            width: args.width,
            quality: args.quality,
            cfgScale: args.cfg_scale,
            seed: args.seed
          },
        }),
      }));
  
      server.sendLoggingMessage({
        level: "info",
        data: "Received response from Bedrock API",
      });
  
      if (!response.body) {
        server.sendLoggingMessage({
          level: "error",
          data: "No response body received from Bedrock",
        });
        throw new McpError(
          ErrorCode.InternalError,
          "No response body received from Bedrock"
        );
      }
  
      const responseBody = JSON.parse(new TextDecoder().decode(response.body)) as BedrockResponse;
  
      if (!responseBody.images || responseBody.images.length === 0) {
        server.sendLoggingMessage({
          level: "error",
          data: "No image data in response",
        });
        throw new McpError(
          ErrorCode.InternalError,
          `No image data in response due to ${responseBody.error}.`
        );
      }
  
      server.sendLoggingMessage({
        level: "info",
        data: "Successfully generated image",
      });
  
      // Return the response in the correct MCP format
      return {
        content: [
          {
            type: "text",
            text: `This is the image generated for your request '${args.prompt}'.`,
          },
          ...responseBody.images.map(image => ({
            type: "image",
            data: image as string,
            mimeType: "image/png",
          })),
          {
            type: "text",
            text: "This is the end of the image generation.",
          }
        ],
      };
    } catch (error) {
      console.error('Error:', error);
  
      // Handle Zod validation errors
      if (error instanceof z.ZodError) {
        throw new McpError(
          ErrorCode.InvalidParams,
          `Invalid parameters: ${error.errors.map(e => e.message).join(", ")}`
        );
      }
  
      // Handle AWS Bedrock errors
      if (error instanceof Error) {
        throw new McpError(
          ErrorCode.InternalError,
          `Failed to generate image: ${error.message}`
        );
      }
  
      // Handle unknown errors
      throw new McpError(
        ErrorCode.InternalError,
        "An unexpected error occurred"
      );
    }
  });

• src/index.ts:65-97 (schema)

  Zod schema for validating input parameters to the generate_image tool, including prompt, dimensions, quality, etc., with refinements for aspect ratio and pixel count.

  const GenerateImageSchema = z.object({
    prompt: z.string().min(1).max(1024, "Prompt must be 1-1024 characters"),
    negativePrompt: z.string().min(1).max(1024, "Negative prompt must be 1-1024 characters").optional(),
    width: z.number().int()
      .min(320, "Width must be at least 320 pixels")
      .max(4096, "Width must be at most 4096 pixels")
      .refine(val => val % 16 === 0, "Width must be divisible by 16")
      .default(1024),
    height: z.number().int()
      .min(320, "Height must be at least 320 pixels")
      .max(4096, "Height must be at most 4096 pixels")
      .refine(val => val % 16 === 0, "Height must be divisible by 16")
      .default(1024),
    quality: z.enum(["standard", "premium"]).default("standard"),
    cfg_scale: z.number().min(1.1).max(10).default(6.5),
    seed: z.number().int().min(0).max(858993459).default(12),
    numberOfImages: z.number().int().min(1).max(5).default(1)
  }).refine(
    (data) => {
      // Check aspect ratio between 1:4 and 4:1
      const ratio = data.width / data.height;
      return ratio >= 0.25 && ratio <= 4;
    },
    "Aspect ratio must be between 1:4 and 4:1"
  ).refine(
    (data) => {
      // Check total pixel count
      return data.width * data.height < 4194304;
    },
    "Total pixel count must be less than 4,194,304"
  );
  
  type GenerateImageInput = z.infer<typeof GenerateImageSchema>;

• src/index.ts:124-172 (registration)

  Registration of the generate_image tool in the ListToolsRequestSchema handler, providing name, description, and JSON input schema.

  server.setRequestHandler(ListToolsRequestSchema, async () => {
    return {
      tools: [
        {
          name: "generate_image",
          description: "Generate image(s) using Amazon Nova Canvas model. The returned data is Base64-encoded string that represent each image that was generated.",
          inputSchema: {
            type: "object",
            properties: {
              prompt: {
                type: "string",
                description: "Text description of the image to generate (1-1024 characters)",
              },
              negativePrompt: {
                type: "string",
                description: "Optional text description of what to avoid in the image (1-1024 characters)",
              },
              width: {
                type: "number",
                description: "Width of the generated image (default: 1024)",
              },
              height: {
                type: "number",
                description: "Height of the generated image (default: 1024)",
              },
              quality: {
                type: "string",
                enum: ["standard", "premium"],
                description: "Quality of the generated image (default: standard)",
              },
              cfg_scale: {
                type: "number",
                description: "How closely to follow the prompt (1.1-10, default: 6.5)",
              },
              seed: {
                type: "number",
                description: "Seed for reproducible generation (0-858993459, default: 12)",
              },
              numberOfImages: {
                type: "number",
                description: "Number of images to generate (1-5, default: 1)",
              },
            },
            required: ["prompt"],
          },
        },
      ],
    };
  });