One question your model's weights can't answer honestly. Install the CWI Discovery Engine into your AI client and get matched tracks with verified ISRCs plus a sync-clearance summary — in one tool call.
Click the button. Cursor opens, pre-fills the server, you approve. Done.
Add to CursorManual alternative: Cursor Settings → MCP → Add custom MCP, type url, value https://cwi-machine-data.hp-ace.workers.dev/mcp.
Our server is remote (no download needed), so there is no bundle to install — paste this into your config file and restart Claude Desktop.
{
"mcpServers": {
"cwi-discovery": {
"url": "https://cwi-machine-data.hp-ace.workers.dev/mcp"
}
}
}
Config file: macOS ~/Library/Application Support/Claude/claude_desktop_config.json · Windows %APPDATA%\Claude\claude_desktop_config.json. Merge the cwi-discovery entry into your existing mcpServers object.
Paste into your terminal. No key, no signup.
claude mcp add --transport http cwi-discovery https://cwi-machine-data.hp-ace.workers.dev/mcp
Then run claude mcp list — cwi-discovery should show connected.
find_tracks_for_brief among the 29.find_tracks_for_brief with this brief:
{"brief": "dark aggressive track for a cyberpunk game trailer"}
QZDA62200554), a why-it-fits reason, and a clearance summary ending with "a match is not a license" plus next steps.No API key is needed — the endpoint is open. Rate limit: 3 write actions per requester per 24h; reads unlimited.
Same matching engine, in your browser, no install: cumulativeweb.com/scout/
Developers: pip install cwi-langchain · PyPI · Full capability map: cwi-discovery.json