TypeScript MCP Server Boilerplate

Instructions

Implementation Reference

• src/index.ts:105-140 (handler)

  The handler function that performs AI image generation from text prompt using Hugging Face InferenceClient with FLUX.1-schnell model, handles errors, converts image to base64.

  async ({ prompt }: { prompt: string }) => {
      try {
          const client = new InferenceClient(config.HF_TOKEN)
          
          const imageBlob = await client.textToImage({
              provider: 'fal-ai',
              model: 'black-forest-labs/FLUX.1-schnell',
              inputs: prompt,
              parameters: { num_inference_steps: 5 }
          }) as unknown as Blob
          
          // Convert Blob to base64
          const arrayBuffer = await imageBlob.arrayBuffer()
          const buffer = Buffer.from(arrayBuffer)
          const base64Data = buffer.toString('base64')
          
          return {
              content: [
                  {
                      type: 'image',
                      data: base64Data,
                      mimeType: 'image/png'
                  }
              ]
          }
      } catch (error) {
          return {
              content: [
                  {
                      type: 'text',
                      text: `이미지 생성 오류: ${error instanceof Error ? error.message : '알 수 없는 오류'}`
                  }
              ]
          }
      }
  }

• src/index.ts:102-104 (schema)

  Zod input schema for the image_generation tool, requiring a 'prompt' string.

  {
      prompt: z.string().describe('Text description of the image to generate')
  },

• src/index.ts:99-141 (registration)

  McpServer.tool registration for 'image_generation' tool including name, description, schema, and handler.

  server.tool(
      'image_generation',
      'Generate an image from a text prompt using AI',
      {
          prompt: z.string().describe('Text description of the image to generate')
      },
      async ({ prompt }: { prompt: string }) => {
          try {
              const client = new InferenceClient(config.HF_TOKEN)
              
              const imageBlob = await client.textToImage({
                  provider: 'fal-ai',
                  model: 'black-forest-labs/FLUX.1-schnell',
                  inputs: prompt,
                  parameters: { num_inference_steps: 5 }
              }) as unknown as Blob
              
              // Convert Blob to base64
              const arrayBuffer = await imageBlob.arrayBuffer()
              const buffer = Buffer.from(arrayBuffer)
              const base64Data = buffer.toString('base64')
              
              return {
                  content: [
                      {
                          type: 'image',
                          data: base64Data,
                          mimeType: 'image/png'
                      }
                  ]
              }
          } catch (error) {
              return {
                  content: [
                      {
                          type: 'text',
                          text: `이미지 생성 오류: ${error instanceof Error ? error.message : '알 수 없는 오류'}`
                      }
                  ]
              }
          }
      }
  )

• src/index.ts:6-8 (schema)

  Configuration schema requiring HF_TOKEN for image generation authentication.

  export const configSchema = z.object({
      HF_TOKEN: z.string().describe('Hugging Face API token for image generation')
  })