TwelveLabs MCP Server

This project is not affiliated with TwelveLabs (the video understanding platform). It is an MCP server for the ElevenLabs Conversational AI API — voice agents, knowledge bases, conversations, and TTS. The name "TwelveLabs" is the project codename only.

An enhanced MCP (Model Context Protocol) server that gives any compatible MCP client direct access to the ElevenLabs Conversational AI API -- including capabilities the official ElevenLabs MCP connector doesn't expose.

What this covers

• Agents -- list, get full config (including LLM model + temperature), update prompt/settings, manage webhooks and file uploads

• Knowledge Base -- list, read full content, add text/URL docs, delete docs

• Conversations -- filter and tag history, inspect transcripts, and rerun post-call analysis/evaluations

• Voices -- list and search available voices

Related MCP server: elevenlabs-mcp-experiments

Prerequisites

• Node.js >= 18

• ElevenLabs API key -- get one at elevenlabs.io

Quick Start

Option A: npx (no clone needed)

ELEVENLABS_API_KEY=your-key npx twelvelabs-mcp-server

Option B: Clone and build

git clone https://github.com/micro-JAY/twelvelabs-mcp-server.git
cd twelvelabs-mcp-server
npm install --include=dev
npm run build

Connect an MCP client

Add an equivalent server entry to your MCP client's configuration. The config file location and surrounding syntax vary by client; this is the standard stdio server definition:

{
  "mcpServers": {
    "twelvelabs": {
      "command": "node",
      "args": ["/path/to/twelvelabs-mcp-server/dist/index.js"],
      "env": {
        "ELEVENLABS_API_KEY": "your-api-key-here"
      }
    }
  }
}

Restart or reconnect your MCP client. You should see twelvelabs in its available tools list.

Use it

Ask your MCP-enabled assistant to use the tools, for example:

"List my agents" "Get the full config for agent_1001kjbh14xce5kbreh55ydf4rae" "Show me the last 10 conversations for that agent" "Get the transcript and analysis for conversation_xyz"

Compatibility

This server works with MCP clients that support the stdio transport and let you configure environment variables for a local server process. Clients that only support remote HTTP MCP servers need a compatible stdio bridge or proxy.

Tool Reference

Agent Tools

Knowledge Base Tools

Conversation Tools

Voice Tools

Security

• Input validation -- all tool parameters are validated with Zod schemas before reaching the API

• HTTPS enforcement -- webhook URLs are validated to use HTTPS only

• Path traversal protection -- document and agent IDs are sanitized to prevent path traversal

• Error sanitization -- internal errors are caught and returned as safe MCP error responses; raw stack traces are never exposed to the client

Development

npm run dev   # watch mode with tsx -- auto-reloads on file changes
npm run build # compile to dist/

Contributing

1. Fork the repository

2. Create a feature branch (git checkout -b feat/my-feature)

3. Commit your changes (git commit -m 'feat: add my feature')

4. Push to the branch (git push origin feat/my-feature)

5. Open a Pull Request

License

MIT -- see LICENSE for details.

Architecture Note

All tools share a single Axios client instance (src/client.ts) initialised from ELEVENLABS_API_KEY at startup. Tool implementations live in src/tools/ -- one file per domain -- and are registered onto the McpServer in src/index.ts. The server communicates over stdio transport (stdin/stdout), which is why all debug logging uses console.error (stderr), not console.log.