💼WorkBuddy Setup Guide

Connect SonicAura MCP to WorkBuddy

Add SonicAura's audio tools to WorkBuddy AI. WorkBuddy uses API Key authentication instead of OAuth, so you will need to paste your SonicAura API key into the config file manually.

⏱ ~3 min setup🔑 API Key auth⚙️ Requires Trust step

Before you start

• WorkBuddy AI 5.6.2 or newer
• A SonicAura account with at least 1 credit
• Your SonicAura API key (from the MCP dashboard)

Step-by-step setup

Install WorkBuddy AI

Download and install the latest version of WorkBuddy AI (5.6.2 or newer). MCP support is available in the desktop app.

Open the MCP config file

WorkBuddy reads MCP servers from a JSON config file in your home directory. Open or create the file below.

macOS

~/.workbuddy-ai/mcp.json

Windows

%USERPROFILE%\.workbuddy-ai\mcp.json

Paste the SonicAura config

WorkBuddy uses API Key authentication (not OAuth). Copy the JSON below and replace YOUR_API_KEY with the key from your MCP dashboard.

{
  "mcpServers": {
    "audio-separator-denoise-enhance": {
      "type": "http",
      "url": "https://api.sonicaurastudio.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Replace YOUR_API_KEY with your key from the MCP dashboard. Keep the space after "Bearer".

Restart WorkBuddy

Fully quit WorkBuddy (not just close the window), then reopen it. The app will reload the MCP config on startup.

Trust the MCP server

Open WorkBuddy → 助手 → 专家·技能·连接器. Find "audio-separator-denoise-enhance" in the list and click 信任 (Trust). Without this step, the connection stays inactive.

Verify the connection

Start a new task and ask WorkBuddy to check your SonicAura credits. If it calls get_my_credits or check_service_status, the setup is complete.

Verify the connection

Start a new task in WorkBuddy and type:

Check my SonicAura credits.

WorkBuddy should call get_my_credits and reply with your remaining credits. If you see a "ToolSearch cannot find audio-separator" error, the config file path or API key is wrong.

A note on WorkBuddy behavior

Heads up: WorkBuddy's Agent may not call check_service_status before trying to upload. This is a client-side behavior, not a SonicAura issue.

If your account has 0 credits, the server will still block the upload at the tool level — you will see a clear "insufficient credits" message. No real work happens until credits are available.

Available tools

check_service_status

Check if the service is ready and how many credits you have.

get_my_credits

Query your remaining separation credits.

get_upload_url

Get a presigned URL to upload large audio files.

upload_audio

Upload small audio files (<1MB) via base64.

submit_separation

Submit a stem separation task (2/4/6 tracks).

check_separation_status

Check task progress and download the separated stems.

Troubleshooting

Why does WorkBuddy need an API key instead of OAuth?

WorkBuddy does not yet implement the MCP OAuth discovery flow. It only supports header-based authentication, so you must paste your SonicAura API key into the "Authorization" header manually.

Where do I find my SonicAura API key?

Sign in at sonicaurastudio.com, then go to /mcp. Your key is generated automatically on first visit and shown in the "Your API Key" field. Copy it and paste it into the WorkBuddy config.

The MCP tools do not appear after restarting WorkBuddy.

Check that the JSON is valid (no trailing commas), that "Authorization" is spelled correctly (capital A), and that the Bearer token has no extra spaces. Then make sure you have clicked 信任 (Trust) on the server in the connectors list.

I see "ToolSearch cannot find audio-separator" errors.

This usually means the MCP server failed to register during startup. Verify the config file path (~/.workbuddy-ai/mcp.json), the URL, and that your API key is active. Then fully quit and reopen WorkBuddy.

WorkBuddy does not call check_service_status before uploading.

This is a known behavior difference between AI clients. Some clients (like WorkBuddy) skip the status check and go straight to uploading. SonicAura server-side checks will still block the request if credits are insufficient, so you will see an error before any real work happens.

Ready to connect another client?

We support Claude Desktop, Cursor, and VS Code too.

View all MCP clients →