Rime MCP

What is Rime MCP?

Rime MCP is a Model Context Protocol (MCP) server that provides text-to-speech capabilities using the Rime API. It downloads synthesized audio and plays it through the system’s native audio player, intended for use with AI agents.

How to use Rime MCP?

Configure it as an MCP tool by adding an entry to your MCP client’s settings with npx rime-mcp and setting the RIME_API_KEY environment variable. Optional environment variables can customize the voice, guidance, and when to speak. The server exposes a speak tool that converts text to speech and plays it.

Key features of Rime MCP

• Exposes a speak tool for text-to-speech
• Uses Rime’s high-quality voice synthesis API
• Plays audio through the system’s native player
• Supports customization via environment variables
• Works out of the box on macOS (uses afplay)
• Open source under the MIT license

Use cases of Rime MCP

• Make a coding agent announce its progress and file changes
• Teach an agent to use Gen Alpha slang and address a specific person
• Switch voices based on the programming language being discussed
• Have an agent speak automatically when finishing a command

FAQ from Rime MCP

What runtime does Rime MCP require?

Node.js 16.x or higher and a working audio output device. On macOS, afplay is used; sample code for Windows and Linux is provided but not tested.

How do I get a Rime API key?

You can create an API key from the Rime Dashboard. Set it as the RIME_API_KEY environment variable.

Which voices are available?

All voice options are listed at https://users.rime.ai/data/voices/all-v2.json. The default voice is cove.

Can I customize how the agent speaks?

Yes. Use RIME_GUIDANCE to describe when and how to speak, RIME_WHO_TO_ADDRESS to set the addressee, RIME_WHEN_TO_SPEAK to control timing, and RIME_VOICE to set a default voice.

Does Rime MCP work on Windows or Linux?

Sample code for Windows and Linux audio playback is included but has not been tested. macOS is fully supported.