MCP Speaker Diarization

You can deploy and use an MCP server that provides persistent speaker identity, emotion-aware transcription, and seamless AI agent integration through a REST API and an MCP endpoint. This server enables you to enroll speakers once, recognize them across conversations, and drive AI workflows with memory and context from multi-party transcripts.

How to use

Connect an MCP client to the server to perform speaker management, real-time transcription, and emotion-aware analysis. The MCP endpoint is designed for AI assistants to query conversations, identify unknown speakers, rename speakers across all past segments, and reprocess conversations when profiles change.

How to install

Prerequisites: you need Python and Docker for flexible deployment.

Option A: Docker deployment (recommended) Start by ensuring you have Docker and NVIDIA tools installed if you plan to use GPU acceleration.

1. Install NVIDIA Container Toolkit (for GPU usage with Docker)

# Ubuntu/Debian
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

MCP client connection (example)

You connect your MCP client to the server at the MCP endpoint. Use the following configuration as a starting point for your Flowise or Claude Desktop integration.

{
  "mcpServers": {
    "speaker-diarization": {
      "url": "http://localhost:8000/mcp",
      "transport": "http"
    }
  }
}

Available tools

list_conversations

Retrieve conversation IDs and metadata for all transcripts.

get_conversation

Fetch a full transcript with speaker labels for a specific conversation.

get_latest_segments

Obtain recent segments across conversations for quick review.

identify_speaker_in_segment

Label an unknown segment with a speaker name and auto-enroll if requested.

rename_speaker

Rename an existing speaker profile and retroactively update all past segments.

list_speakers

List all enrolled speaker profiles with their metadata.

delete_speaker

Remove a speaker profile from the system.

delete_all_unknown_speakers

Clear all Unknown_* speakers from the dataset.

update_conversation_title

Set or update the title of a conversation.

reprocess_conversation

Re-run recognition on a conversation using updated profiles.

search_conversations_by_speaker

Find all conversations where a specific speaker appears.