ElevenLabs MCP Server

Instructions

Implementation Reference

• elevenlabs_mcp/server.py:323-393 (handler)

  The handler function decorated with @mcp.tool that implements the speech_to_text tool logic. It handles audio input, calls the ElevenLabs speech-to-text API (scribe_v1 model), supports diarization, formats the transcript, and manages output based on configuration.
  @mcp.tool( description=f"""Transcribe speech from an audio file. When save_transcript_to_file=True: {get_output_mode_description(output_mode)}. When return_transcript_to_client_directly=True, always returns text directly regardless of output mode. ⚠️ COST WARNING: This tool makes an API call to ElevenLabs which may incur costs. Only use when explicitly requested by the user. Args: file_path: Path to the audio file to transcribe language_code: ISO 639-3 language code for transcription. If not provided, the language will be detected automatically. diarize: Whether to diarize the audio file. If True, which speaker is currently speaking will be annotated in the transcription. save_transcript_to_file: Whether to save the transcript to a file. return_transcript_to_client_directly: Whether to return the transcript to the client directly. output_directory: Directory where files should be saved (only used when saving files). Defaults to $HOME/Desktop if not provided. Returns: TextContent containing the transcription or MCP resource with transcript data. """ ) def speech_to_text( input_file_path: str, language_code: str | None = None, diarize: bool = False, save_transcript_to_file: bool = True, return_transcript_to_client_directly: bool = False, output_directory: str | None = None, ) -> Union[TextContent, EmbeddedResource]: if not save_transcript_to_file and not return_transcript_to_client_directly: make_error("Must save transcript to file or return it to the client directly.") file_path = handle_input_file(input_file_path) if save_transcript_to_file: output_path = make_output_path(output_directory, base_path) output_file_name = make_output_file("stt", file_path.name, "txt") with file_path.open("rb") as f: audio_bytes = f.read() if language_code == "" or language_code is None: language_code = None transcription = client.speech_to_text.convert( model_id="scribe_v1", file=audio_bytes, language_code=language_code, enable_logging=True, diarize=diarize, tag_audio_events=True, ) # Format transcript with speaker identification if diarization was enabled if diarize: formatted_transcript = format_diarized_transcript(transcription) else: formatted_transcript = transcription.text if return_transcript_to_client_directly: return TextContent(type="text", text=formatted_transcript) if save_transcript_to_file: transcript_bytes = formatted_transcript.encode("utf-8") # Handle different output modes success_message = f"Transcription saved to {file_path}" return handle_output_mode( transcript_bytes, output_path, output_file_name, output_mode, success_message, ) # This should not be reached due to validation at the start of the function return TextContent(type="text", text="No output mode specified")
• elevenlabs_mcp/server.py:323-340 (registration)

  The @mcp.tool decorator registers the speech_to_text tool with MCP, including the tool description and argument details.
  @mcp.tool( description=f"""Transcribe speech from an audio file. When save_transcript_to_file=True: {get_output_mode_description(output_mode)}. When return_transcript_to_client_directly=True, always returns text directly regardless of output mode. ⚠️ COST WARNING: This tool makes an API call to ElevenLabs which may incur costs. Only use when explicitly requested by the user. Args: file_path: Path to the audio file to transcribe language_code: ISO 639-3 language code for transcription. If not provided, the language will be detected automatically. diarize: Whether to diarize the audio file. If True, which speaker is currently speaking will be annotated in the transcription. save_transcript_to_file: Whether to save the transcript to a file. return_transcript_to_client_directly: Whether to return the transcript to the client directly. output_directory: Directory where files should be saved (only used when saving files). Defaults to $HOME/Desktop if not provided. Returns: TextContent containing the transcription or MCP resource with transcript data. """ )
• elevenlabs_mcp/server.py:75-154 (helper)

  Supporting helper function used by speech_to_text to format transcripts with speaker labels when diarization is enabled.
  def format_diarized_transcript(transcription) -> str: """Format transcript with speaker labels from diarized response.""" try: # Try to access words array - the exact attribute might vary words = None if hasattr(transcription, "words"): words = transcription.words elif hasattr(transcription, "__dict__"): # Try to find words in the response dict for key, value in transcription.__dict__.items(): if key == "words" or ( isinstance(value, list) and len(value) > 0 and ( hasattr(value[0], "speaker_id") if hasattr(value[0], "__dict__") else ( "speaker_id" in value[0] if isinstance(value[0], dict) else False ) ) ): words = value break if not words: return transcription.text formatted_lines = [] current_speaker = None current_text = [] for word in words: # Get speaker_id - might be an attribute or dict key word_speaker = None if hasattr(word, "speaker_id"): word_speaker = word.speaker_id elif isinstance(word, dict) and "speaker_id" in word: word_speaker = word["speaker_id"] # Get text - might be an attribute or dict key word_text = None if hasattr(word, "text"): word_text = word.text elif isinstance(word, dict) and "text" in word: word_text = word["text"] if not word_speaker or not word_text: continue # Skip spacing/punctuation types if they exist if hasattr(word, "type") and word.type == "spacing": continue elif isinstance(word, dict) and word.get("type") == "spacing": continue if current_speaker != word_speaker: # Save previous speaker's text if current_speaker and current_text: speaker_label = current_speaker.upper().replace("_", " ") formatted_lines.append(f"{speaker_label}: {' '.join(current_text)}") # Start new speaker current_speaker = word_speaker current_text = [word_text.strip()] else: current_text.append(word_text.strip()) # Add final speaker's text if current_speaker and current_text: speaker_label = current_speaker.upper().replace("_", " ") formatted_lines.append(f"{speaker_label}: {' '.join(current_text)}") return "\n\n".join(formatted_lines) except Exception: # Fallback to regular text if something goes wrong return transcription.text @mcp.resource("elevenlabs://{filename}")