ElevenLabs MCP Server

Overview

The ElevenLabs MCP Server connects AI assistants to ElevenLabs. The supported implementation is the hosted remote server at https://api.elevenlabs.io/v1/mcp, which exposes agent management and text-to-speech generation through a standardised MCP interface; the local server in the GitHub repository is deprecated.

Key Features

Available Tools

The hosted server exposes agent-management tools and text-to-speech generation. ElevenLabs' documentation describes the capability areas — creating and updating agents, listing agents and their configuration, reviewing conversations and transcripts, managing tools and knowledge base documents, working with tests and deployments, and generating speech audio from text — but does not publish a fixed tool-name list on that page. Tool names and arguments are discovered from the server when your client connects.

Installation

Point your client at the hosted server URL and complete OAuth:

{
  "mcpServers": {
    "elevenlabs": {
      "url": "https://api.elevenlabs.io/v1/mcp"
    }
  }
}

For Claude Desktop, open Settings → Connectors, search for ElevenLabs, select Connect, and finish the OAuth flow. For isolated data-residency workspaces, use the region URL instead:

No API key is copied into the client; access is scoped by the OAuth permissions you approve.

Output Configuration

The hosted server returns generated speech as a short-lived download link. The ELEVENLABS_MCP_OUTPUT_MODE and ELEVENLABS_MCP_BASE_PATH variables belong to the deprecated local package only and do not apply to the hosted server.

Example Use Cases

• Agent Management: Create agents from a description, then update prompts, voices, languages, and first messages
• Conversation Review: Read recent conversations and full transcripts, and search across them
• Cost Estimation: Estimate an agent's expected LLM usage and cost before changing models
• Testing: Create, run, and review tests for your agents
• Speech Generation: Generate speech audio from text and return a short-lived download link

Security

• OAuth authentication — no API keys are copied into the client
• Access is scoped by the permissions you approve and limited to the workspace you sign in with
• Tool controls in your MCP client can require approval before each call
• Deleting an agent is destructive; restrict write access to workspace members who need it
• All communication encrypted via HTTPS

Sources

• https://github.com/elevenlabs/elevenlabs-mcp
• https://elevenlabs.io/docs/agents-platform/operate/hosted-mcp