ElevenLabs MCP Server – AI Audio Tools
The official ElevenLabs MCP server exposes ElevenLabs audio and conversational AI APIs to MCP-compatible clients. AI agents can use it to generate speech, clone or design voices, transcribe audio, create sound effects, and build or manage voice-agent workflows using an ElevenLabs API key.
Overview
ElevenLabs MCP Server is the official Model Context Protocol server maintained by ElevenLabs. It runs as a local MCP server and lets tools such as Claude Desktop, Cursor, Windsurf, OpenAI Agents, and other MCP clients call ElevenLabs audio APIs through structured tools rather than ad hoc scripts.
The server is useful when an AI assistant needs to produce or process audio as part of a workflow. Instead of only describing how to use ElevenLabs, the assistant can call MCP tools that forward requests to ElevenLabs services using the configured account credentials.
What the MCP server enables
The official repository describes the server as enabling interaction with ElevenLabs Text to Speech and audio processing APIs. Documented examples include generating speech, cloning voices, transcribing audio, creating voice variations, converting voice recordings, creating soundscapes, and building conversational AI voice agents.
The server can return generated outputs as local files, MCP resources, or both, depending on ELEVENLABS_MCP_OUTPUT_MODE. Optional file handling settings include ELEVENLABS_MCP_BASE_PATH for relative file operations and output mode controls for environments where direct disk access is not ideal.
When to use it
Use this MCP server when an AI agent needs direct access to ElevenLabs audio capabilities from an MCP client. Practical workflows include creating spoken versions of text, prototyping character voices, transcribing meeting or interview audio, preparing voice-agent demos, generating audio assets, or testing conversational AI behaviors from an IDE or desktop assistant.
It is also useful for developers who want a local, open-source bridge between an MCP client and ElevenLabs cloud APIs without writing separate integration code for every audio task.
Connection and authentication
The documented quickstart uses a local stdio MCP server launched with uvx and the elevenlabs-mcp package. Authentication is configured with the ELEVENLABS_API_KEY environment variable. Other MCP clients can install the package with pip install elevenlabs-mcp and run python -m elevenlabs_mcp --api-key={{PUT_YOUR_API_KEY_HERE}} --print to generate client configuration.
Optional environment variables include ELEVENLABS_MCP_BASE_PATH, ELEVENLABS_MCP_OUTPUT_MODE, and ELEVENLABS_API_RESIDENCY. Data residency is documented as an enterprise-only feature.
Key considerations
ElevenLabs credits are required to use the audio tools, even though the MCP server itself is open source. Some API operations, such as voice design and audio isolation, can take longer to complete and may appear as timeouts in development inspectors even when the task finishes. Keep the API key in environment variables or a client secret store, avoid committing it to configuration files, and choose an output mode that matches the client environment and data-handling requirements.
Supported Transports
stdio
Command: uvx
Args:
elevenlabs-mcp
Frequently Asked Questions
- When should an AI agent use the ElevenLabs MCP server?
- Use it when an AI agent needs to generate, transform, or analyze audio through ElevenLabs from an MCP client. Common documented workflows include generating speech, creating custom voice variations, transcribing audio, converting recordings, creating soundscapes, and building conversational AI voice agents.
- What does the ElevenLabs MCP server add to an AI agent's capabilities?
- It gives the agent tool-based access to ElevenLabs Text to Speech and audio processing APIs, so the agent can perform audio tasks directly instead of only explaining how to call the ElevenLabs API manually.
- What can the AI agent access or manage through this server?
- The official documentation and repository describe access to speech generation, voice cloning or design workflows, speech-to-text transcription, soundscape generation, voice conversion, and conversational AI voice-agent workflows. Use of these capabilities depends on the configured ElevenLabs account, API key, credits, and product availability.
- How is authentication configured for this MCP server?
- The documented Claude Desktop configuration runs the local server with `uvx elevenlabs-mcp` and sets `ELEVENLABS_API_KEY` in the MCP server environment. Other clients can install the package and generate configuration with the Python module command shown in the official README. Never place a real ElevenLabs API key directly in a shared YAML file.
- Which transport should be used for this MCP server?
- Use the documented local stdio transport. The official quickstart starts the MCP server as a local command with `uvx` and connects MCP clients to that process. This YAML does not include hosted HTTP or SSE transports because the official ElevenLabs MCP repository documents the local command-based setup for client integration.