ElevenLabs MCP Server

This skill provides a comprehensive suite of 24 audio tools from ElevenLabs, exposed through a standard Model Context Protocol (MCP) server. When this skill is used, it starts the server, allowing the agent to directly call powerful audio generation and manipulation capabilities.

When to Use

This skill should be loaded by an agent platform (like OpenClaw) at startup. It is not meant to be called directly in a conversation but rather to provide the tools that other skills or the agent itself will use.

• For Agent Developers: Include this skill in your agent's default toolkit to give it a full range of audio abilities.
• For Users: You don't need to call this skill directly. Instead, use a skill like audio-conductor which will intelligently use the tools provided by this server.

How It Works

• Server Initialization: When the skill is loaded, it executes the tools/start_server.sh script.
• Environment Setup: The script sets the ELEVENLABS_API_KEY from the agent's environment variables.
• Start MCP Server: It launches the elevenlabs-mcp server process in the background on a designated port (e.g., 8124).
• Tool Registration: The server automatically registers all 24 of its tools (e.g., text_to_speech, compose_music) with the agent's MCP client.
• Ready for Use: The agent can now see and call these tools directly as part of its decision-making process.

Provided Tools (Partial List)

| Tool Name | Description | | :--- | :--- | | text_to_speech | Converts text to speech. | | text_to_sound_effects | Generates a sound effect from a description. | | compose_music | Creates a full music track from a prompt. | | voice_clone | Clones a voice from audio samples. | | speech_to_text | Transcribes audio to text. | | search_voices | Searches for available voices. |

For a full list, see the references/tool_list.md.

Usage in Other Skills

An orchestrator skill like audio-conductor would use these tools like this (conceptual example):

# Inside audio-conductor logic
if audio_type == 'sfx':
    # Directly call the MCP tool provided by this server
    result = await mcp.call_tool('text_to_sound_effects', {'text': prompt})

References

• tool_list.md: A complete list and description of all 24 tools provided by the server.
• start_server.sh: The script responsible for launching the MCP server.