claude-code - 💡(How to fix) Fix ElevenLabs MCP connector: Text_To_Speech tool schema missing required `text` parameter

Summary

The hosted ElevenLabs MCP connector exposes Text_To_Speech (and three sibling TTS tools) without the required text parameter in its JSON schema. Every call returns 404 Not Found from the upstream ElevenLabs API because no text body is forwarded.

Affected tools

All four return the same error:

• mcp__elevenlabs__Text_To_Speech
• mcp__elevenlabs__Text_To_Speech_Streaming
• mcp__elevenlabs__Text_To_Speech_Streaming_With_Timestamps
• mcp__elevenlabs__Text_To_Speech_With_Timestamps

Each exposes only: voice_id, output_format, optimize_streaming_latency, enable_logging, xi-api-key. None expose text.

The dialogue variants likely have the analogous gap (missing inputs array):

• mcp__elevenlabs__Text_To_Dialogue_Multi-Voice_
• mcp__elevenlabs__Text_To_Dialogue_Multi-Voice_Streaming
• mcp__elevenlabs__Text_To_Dialogue_Streaming_With_Timestamps
• mcp__elevenlabs__Text_To_Dialogue_With_Timestamps

Reproduction

mcp__elevenlabs__Text_To_Speech(
  voice_id="EXAVITQu4vr4xnSDxMaL",  # Sarah - premade, verified via List_Voices
  output_format="mp3_44100_128"
)

Result:

MCP error -32603: API request failed:
Request failed with status code 404 - {"detail":"Not Found"}

Voice ID validity was verified via List_Voices (premade, english, american). The 404 is from the upstream ElevenLabs API rejecting a body that lacks text, not from a missing voice.

Expected

Schema should declare text as a required string parameter and forward it as the JSON body field to POST /v1/text-to-speech/{voice_id} per the ElevenLabs REST API spec: https://elevenlabs.io/docs/api-reference/text-to-speech/convert

A correct schema would look like:

{
  "required": ["voice_id", "text"],
  "properties": {
    "voice_id": {"type": "string"},
    "text": {"type": "string"},
    "model_id": {"type": "string"},
    "voice_settings": {"type": "object"},
    "output_format": {"type": "string"},
    ...
  }
}

Likely also missing: model_id, voice_settings, language_code, pronunciation_dictionary_locators, seed, previous_text, next_text, previous_request_ids, next_request_ids — all standard fields on the ElevenLabs TTS endpoint.

Impact

The TTS tools are unusable. Workarounds (shell out to curl against the REST API, or use OS-level say) defeat the purpose of having the connector.

Environment

• Claude Code CLI (Opus 4.7, 1M context)
• macOS 25.5.0 (Darwin)
• ElevenLabs MCP server id: aabeb1b7-... (hosted connector)