Voxtral TTS MCP

What Voxtral TTS MCP MCP server does

Voxtral TTS MCP MCP server gives an MCP client access to reference information about Voxtral TTS. It does not provide a speech-generation API through the listed tools; instead, it exposes the canonical information available from the Voxtral TTS site in a format an agent can retrieve and use.

The server is read-only and has no API-key requirement. Its documented information covers voices, supported languages, TTS modes, frequently asked questions, and official URLs. The underlying Voxtral TTS service is described as supporting voice cloning, expressive styles, multilingual speech, hosted API access, and self-hosted model weights, but this MCP package presents that information rather than performing inference itself.

How it works

The server communicates with MCP clients over standard input and output. After it is added to a client configuration, an agent can call the available tools without supplying parameters. Results are returned as text or Markdown, which makes them suitable for answering questions about available voices, configuration, and official resources.

Two MCP resources provide additional structured reference surfaces: site://voxtraltts/voices covers voices, languages, and TTS modes; site://voxtraltts/faq provides a short FAQ based on public site metadata; and site://voxtraltts/links contains canonical URLs. The package also declares prompts for summarizing Voxtral TTS and planning a read-aloud workflow.

Setup and configuration

Smithery installation is available with the MCP CLI. The README provides client-specific examples for Claude Desktop, Cursor, Windsurf, and Continue. From source, clone the repository, install dependencies with pnpm install, and configure the client to launch src/index.mjs with Node.

A source-based configuration needs the absolute path to that entry file. The documented development command starts the server over stdio. MCP Inspector can launch the same Node entry point for debugging. No environment variables or API credentials are documented as required.

Tools and capabilities

• list_voices: returns the canonical voice and TTS configuration exposed on the Voxtral TTS site. It accepts no input and returns text or Markdown.
• get_official_links: returns the canonical official links, including the website, support contact, and documentation links when available. It also accepts no input and returns text or Markdown.

The tools are useful for agents that need to identify supported voice options or direct users to authoritative Voxtral TTS resources. They should not be treated as audio-generation commands.

Limitations and notes

Voxtral TTS MCP is read-only. The documented interface contains lookup tools, resources, and prompts, but no tool for submitting text, creating audio, cloning a voice, or managing an account. The server therefore fits research, planning, and link-discovery workflows more directly than applications that need live TTS output.

The package is described as zero-config and does not list required environment variables. Its returned material is tied to the canonical Voxtral TTS information surface, so client behavior depends on how the MCP host presents tool, resource, and prompt results. For actual synthesis details or service access, use the official links returned by the server.