SongPilot MCP Server

Model Context Protocol (MCP) server for integrating SongPilot AI agents with Claude Desktop and other MCP clients.

Overview

SongPilot MCP is a client-side package that enables AI assistants like Claude to interact with your SongPilot workspace. It runs locally on your machine and communicates with SongPilot's backend services.

Installation

From PyPI (Recommended)

pip install songpilot-mcp

From Source

git clone https://github.com/songpilotai/songpilot-mcp.git
cd songpilot-mcp
pip install -r requirements.txt
pip install -e .

Configuration

Create a .env file in your working directory:

SONGPILOT_API_KEY=sp_your_api_key_here
SONGPILOT_WORKSPACE_ID=your_workspace_id
SONGPILOT_MCP_BASE_URL=https://mcp.songpilot.ai

Or set environment variables directly:

export SONGPILOT_API_KEY=sp_your_api_key_here
export SONGPILOT_WORKSPACE_ID=your_workspace_id

Getting Your API Key

1. Log into your SongPilot account
2. Go to Settings > API Keys
3. Generate a new key (format: sp_...)
4. Copy your workspace ID from the URL or settings page

Usage with Claude Desktop

Add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "songpilot": {
      "command": "songpilot-mcp",
      "env": {
        "SONGPILOT_API_KEY": "sp_your_api_key",
        "SONGPILOT_WORKSPACE_ID": "your_workspace_id"
      }
    }
  }
}

On macOS, the config file is located at: ~/Library/Application Support/Claude/claude_desktop_config.json

On Windows: %APPDATA%\Claude\claude_desktop_config.json

What You Can Do

Once connected, you can ask Claude to:

• Create artwork for songs and albums
• Plan release strategies
• Write artist bios and social media content
• Analyze your music catalog
• Get career management advice

Example prompts:

"Create artwork for my song 'Midnight Dreams'"
"Help me plan a release strategy for my new album"
"Write a bio for my artist profile"
"Which of my songs are ready for release?"

Architecture

This is a client-side MCP implementation:

• Runs on your local machine (not on SongPilot servers)
• Communicates via stdio with Claude Desktop
• Makes HTTP API calls to SongPilot services
• No Docker or Kubernetes required

Development

Setup

make venv          # Create virtual environment
make install-dev   # Install dependencies

Commands

make format        # Format code with black and isort
make lint          # Run ruff linter and format checks
make test          # Run pytest with coverage
make build         # Build package for distribution
make publish       # Publish to PyPI (requires token)

Testing

# Run all tests
pytest

# With coverage
pytest --cov=songpilot_mcp --cov-report=html

# Specific test file
pytest tests/unit/test_client.py -v

Environment Variables

Variable	Required	Default	Description
SONGPILOT_API_KEY	Yes	-	Your SongPilot API key (sp_...)
SONGPILOT_WORKSPACE_ID	Yes	-	Your workspace ID
SONGPILOT_MCP_BASE_URL	No	https://mcp.songpilot.ai	API base URL
LOG_LEVEL	No	INFO	Logging level (DEBUG, INFO, WARNING, ERROR)

Troubleshooting

"Authentication failed"

• Verify your API key starts with sp_
• Check that the key is active in your SongPilot settings
• Ensure the workspace ID is correct

"Workspace not found"

• Confirm you're using the correct workspace ID
• Verify you have access to the workspace

Connection issues

• Check your internet connection
• Verify SONGPILOT_MCP_BASE_URL is correct (default: https://mcp.songpilot.ai)
• Check SongPilot status page for service outages

License

MIT License - see LICENSE file for details.

Support

• Documentation: https://docs.songpilot.ai
• Issues: https://github.com/songpilotai/songpilot-mcp/issues
• Email: support@songpilot.ai