Audio Studio MCP Lets Claude Generate Stable Audio 3 Locally for Free

Audio Studio MCP is an open-source server that lets Claude Code, Claude Desktop, Cursor, and other MCP clients generate and edit audio with Stable Audio 3 on local hardware. Model Context Protocol gives AI clients a standard way to call external tools. Users describe a track or sound effect in plain language, and the server writes WAV, OGG, or FLAC files for game engines, video editors, and other production tools.

After local setup, the audio server needs no API key and adds no per-generation fee. The machine supplies the compute, and audio synthesis does not call a hosted music API. The repository passed 500 GitHub stars in under two weeks and now includes a browser interface for generation and waveform editing outside an MCP conversation.

Four tools cover the path from prompt to export

Audio Studio registers four MCP tools, giving agents a compact interface for creating, revising, and preparing clips:

• generate_audio creates one to four background-music, instrument-loop, stem, or sound-effect candidates. It accepts BPM and structured sound-effect fields.
• edit_audio restyles a complete clip, regenerates a selected time range with inpainting, or extends a track. Each operation can return one to four candidates.
• process_audio trims, converts to mono, resamples, adjusts loudness, creates seamless loops, adds fades, normalizes peaks, and exports files.
• unload_models immediately releases model memory.

Audio Studio also translates broad requests into the prompt syntax Stable Audio 3 expects. A request such as make it sound like a creaky door receives the appropriate model prefix through the audio_type option, which accepts