An MCP server that provides seamless integration between Fish Audio's Text-to-Speech API and LLMs such as Claude, enabling natural language-driven speech synthesis. It supports two primary tools: fish_audio_tts, which generates speech from text with configurable voice references, streaming, format, bitrate, latency, and playback options; and fish_audio_list_references, which lists all configured voice references. The server allows selecting voices by ID, name, or tags and supports both single-reference and multi-reference configurations via environment variables. Output formats include MP3, WAV, PCM, and Opus, with configurable MP3 bitrate. Configuration is environment-variable driven, including API keys, model IDs, references, default references, output formats, streaming modes, latency, and output directory. The MCP server aims for low-latency streaming and real-time playback, and it emphasizes easy integration with any MCP-compatible client. It includes API documentation, error handling, troubleshooting guidance, and a clear project structure with a dedicated Fish Audio API client and TTS tool, enabling maintainable development and reliable text-to-speech capabilities for LLM-powered workflows.