Elevenlabs Server

Official ElevenLabs MCP server for text-to-speech, voice cloning, audio isolation, and sound effect generation within Claude Code.

When to Use This MCP Server

Connect this server when you need to:

• Generate natural-sounding speech from text using dozens of AI voice models directly from your development workflow
• Clone voices or design custom voice profiles for application prototyping and testing
• Process audio files for transcription, isolation, or sound effect generation without leaving your IDE

Consider alternatives when:

• You only need basic system TTS and do not require high-fidelity or custom voices
• Your project has no audio output requirements and adding an audio MCP server would be unnecessary overhead

Quick Start

Configuration

Copy
name: elevenlabs-server
type: mcp
category: audio

Example Connection

Copy
claude mcp add elevenlabs-server -- uvx elevenlabs-mcp

Available Tools

text-to-speech:       Convert text into natural speech using any ElevenLabs voice
voice-clone:          Create a cloned voice from uploaded audio samples
voice-design:         Design a new synthetic voice with specific characteristics
audio-isolation:      Separate vocals from background audio in a file
sound-effects:        Generate sound effects from text descriptions
transcribe:           Transcribe audio files into text with timestamps

Core Concepts

ElevenLabs Integration Overview

Aspect	Details
Protocol	MCP over stdio transport
Authentication	API key via ELEVENLABS_API_KEY environment variable
Output Modes	File-based output or base64 inline responses
Supported Formats	MP3, WAV, PCM, OGG output formats
Voice Library	Access to hundreds of pre-built and community voices
Rate Limits	Tied to your ElevenLabs subscription tier

Audio Pipeline Architecture

┌─────────────┐     ┌──────────────────┐     ┌────────────────┐
│ Claude Code  │────▶│  ElevenLabs MCP  │────▶│ ElevenLabs API │
│  (Client)    │◀────│    Server        │◀────│   Cloud        │
└─────────────┘     └──────────────────┘     └────────────────┘
                           │
                    ┌──────┴──────┐
                    │  Local File │
                    │   Output    │
                    └─────────────┘

Configuration

Parameter	Type	Default	Description
ELEVENLABS_API_KEY	string	required	Your ElevenLabs API key for authentication
ELEVENLABS_MCP_BASE_PATH	string	~/Desktop	Directory where generated audio files are saved
ELEVENLABS_MCP_OUTPUT_MODE	string	files	Output mode: files saves to disk, inline returns base64
voice_id	string	varies	Specific voice identifier to use for TTS generation
model_id	string	eleven_multilingual_v2	ElevenLabs model for speech synthesis quality

Best Practices

1. Store API Keys Securely - Never hardcode your ELEVENLABS_API_KEY in project files. Use environment variables or a secrets manager, and add .env files to .gitignore to prevent accidental exposure.

2. Choose the Right Output Mode - Use files mode during development to inspect generated audio easily on disk. Switch to inline mode for automated pipelines where files would accumulate unnecessarily.

3. Select Voices by Use Case - Browse the ElevenLabs voice library to pick voices that match your application context. Clone voices only when pre-built options do not meet your specific tonal or branding requirements.

4. Monitor Usage Quotas - ElevenLabs charges by character count. Track your consumption through their dashboard and set alerts before hitting tier limits, especially during batch generation workflows.

5. Leverage Audio Isolation for Clean Input - When working with user-uploaded audio for voice cloning, run audio isolation first to strip background noise. Cleaner samples produce significantly better cloned voice quality.

Common Issues

1. Authentication Failures - Verify your ELEVENLABS_API_KEY is correctly set in the environment. Regenerate the key from the ElevenLabs dashboard if it has expired or been revoked.

2. Empty Audio Output - Check that ELEVENLABS_MCP_BASE_PATH points to a writable directory. The server will silently fail if it cannot write to the specified path.

3. Rate Limit Errors - If you receive 429 responses, you have exceeded your subscription quota. Upgrade your plan or wait for the rate limit window to reset before retrying.