Multimodal & Audio Content

Beyond Text and Images

The 2025-2026 spec expansions added support for audio content blocks, enabling MCP servers to interface with voice analysis, transcription, and Text-to-Speech (TTS) APIs.

Audio Content Blocks

// Returning audio from a TTS tool:
server.tool("text_to_speech", "Convert text to speech",
  { text: z.string(), voice: z.string().optional() },
  async ({ text, voice }) => {
    const audioBuffer = await ttsEngine.synthesize(text, voice);
    return {
      content: [{
        type: "audio",
        data: audioBuffer.toString("base64"),
        mimeType: "audio/wav"
      }]
    };
  }
);

Content Block Types (2026)