Tavus MCP Server

Instructions

Implementation Reference

• src/index.ts:1045-1053 (handler)

  The main handler function that executes the generate_speech tool logic. It makes a POST request to the '/speech' API endpoint with the provided arguments and returns the response as formatted JSON.

  private async generateSpeech(args: any) {
    const response = await this.axiosInstance.post('/speech', args);
    return {
      content: [{
        type: 'text',
        text: JSON.stringify(response.data, null, 2),
      }],
    };
  }

• src/index.ts:603-629 (schema)

  The input schema definition for the generate_speech tool, defining the required parameters: replica_id (required), script (required), speech_name (optional), and callback_url (optional).

  // Speech
  {
    name: 'generate_speech',
    description: 'Generate speech audio from text using a replica',
    inputSchema: {
      type: 'object',
      properties: {
        replica_id: {
          type: 'string',
          description: 'Replica to use for speech generation',
        },
        script: {
          type: 'string',
          description: 'Text script to convert to speech',
        },
        speech_name: {
          type: 'string',
          description: 'Name for the generated speech',
        },
        callback_url: {
          type: 'string',
          description: 'URL to receive completion callback',
        },
      },
      required: ['replica_id', 'script'],
    },
  },

• src/index.ts:750-751 (registration)

  The switch-case routing that maps the 'generate_speech' tool name to its handler method generateSpeech().

  case 'generate_speech':
    return await this.generateSpeech(request.params.arguments);