voicevox

This MCP server acts as a bridge to expose VOICEVOX-compatible text-to-speech capabilities through the Model Context Protocol (MCP). It enables voice synthesis using VOICEVOX engines (VOICEVOX, AivisSpeech, COEIROINK) via MCP clients, allowing agents and tools (e.g., Claude 3.7 agents) to request speech synthesis in a standardized way. To use it, start the MCP server via the configured command (typically through npx as shown in the config). The server will communicate with a locally running VOICEVOX engine API (by default at http://localhost:50021) and expose endpoints compatible with MCP so clients can request speech synthesis by providing text and speaker settings. You can adjust the targeted VOICEVOX engine URL and speaker by setting VOICEVOX_API_URL and VOICEVOX_SPEAKER_ID in the environment or your MCP runtime configuration. Once running, clients connect to the MCP server in your environment and invoke the appropriate MCP micro-tools to synthesize speech for given prompts or texts.

Prerequisites:

• Node.js 18 or newer
• npm (comes with Node.js)
• A VOICEVOX engine instance running locally (e.g., VOICEVOX Engine at http://localhost:50021)

Installation steps:

1. Clone the repository

git clone https://github.com/Dosugamea/voicevox-mcp-server.git
cd voicevox-mcp-server

2. Install dependencies

npm install

3. Configure environment

• Copy the example env and edit as needed (VOICEVOX_API_URL, VOICEVOX_SPEAKER_ID)

cp .env_example .env

4. Run or deploy via MCP integration

• If using the MCP config in this document, you can start the server through your MCP orchestration (e.g., via npx as shown in the mcp_config).

Notes:

• Ensure the VOICEVOX engine is accessible at the URL you provide in VOICEVOX_API_URL.
• The server depends on a running VOICEVOX engine; start it separately if needed.

Environment variables:

• VOICEVOX_API_URL: Base URL for the VOICEVOX engine API (default http://localhost:50021)
• VOICEVOX_SPEAKER_ID: Speaker/model ID to use for synthesis (default 1)

Common issues:

• If speech synthesis fails, verify VOICEVOX Engine is running and accessible from the MCP server host.
• Ensure the VOICEVOX engine supports the chosen speaker; adjust VOICEVOX_SPEAKER_ID accordingly by querying the VOICEVOX Engine /speakers endpoint.
• For Windows, ensure Node.js 18+ is installed and VLC (if used for playback) is accessible.
• For Docker environments, confirm network bindings and VOICEVOX_ENGINE URL are correctly exposed to the container.

Troubleshooting tips:

• Check the MCP client logs for any endpoint or protocol mismatch.
• Validate that the environment variables are loaded by the MCP runtime before starting the server.