Sonicaura MCP

What sonicaura-mcp does

The sonicaura-mcp MCP server exposes SonicAura’s audio-processing functions through MCP. It can split tracks into two, four, or six stems; remove noise from music; clean speech recordings; and add reverb using predefined room and space presets.

Stem separation supports these layouts:

• Two stems: vocals and instrumental
• Four stems: vocals, drums, bass, and other
• Six stems: vocals, guitar, bass, drums, piano, and other

The dedicated extract_vocals tool provides a shortcut for two-stem separation and returns isolated vocals plus the instrumental. This fits workflows such as karaoke preparation, a cappella extraction, covers, and vocal practice.

How it works

The server is hosted at https://api.sonicaurastudio.com/mcp and uses Streamable HTTP transport with OAuth 2.0 authentication. An MCP client sends audio-processing requests to SonicAura rather than running separation or enhancement models locally.

Audio can enter the workflow in two ways. Files under 1 MB can be sent through upload_audio as base64 data. Larger files use get_upload_url, which returns a presigned upload URL for direct transfer to R2 storage. The returned source key is then supplied to process_audio or extract_vocals.

Processing is asynchronous. After submitting a task, use check_task_status to poll its progress and result state. list_audio_tools reports available task types and metadata, while check_service_status checks service availability and credits. get_my_credits reports the current user’s remaining credits.

Setup and configuration

Add the remote endpoint to an MCP client configuration using the HTTP transport:

{
  "mcpServers": {
    "audio-process": {
      "type": "http",
      "url": "https://api.sonicaurastudio.com/mcp"
    }
  }
}

The sonicaura-mcp MCP server uses OAuth 2.0 rather than a documented local environment-variable configuration. Processing consumes credits, so an account and credits from SonicAura are required for audio tasks.

Tools and capabilities

• list_audio_tools: list supported processing types and metadata.
• get_upload_url: obtain a presigned URL for files larger than 1 MB.
• upload_audio: upload files smaller than 1 MB using base64.
• process_audio: submit denoising, reverb, or stem-separation jobs.
• extract_vocals: request vocals and instrumental stems as an easy two-stem operation.
• check_task_status: poll a submitted processing task.
• get_my_credits: inspect remaining credits.
• check_service_status: check whether the audio service is available.

Supported task types include music_denoise, denoise, reverb, 2stems, 4stems, and 6stems. Reverb presets include vocal booth, studio room, bathroom, concert hall, and church settings.

Limitations and notes

The server depends on SonicAura’s remote service, network access, OAuth authentication, and available processing credits. upload_audio is limited to files under 1 MB; larger files must use the presigned-URL path. submit_separation and check_separation_status are deprecated; use process_audio and check_task_status instead. The sonicaura-mcp MCP server does not document local model execution or offline processing.