Cheesecube MCP Server
# Cheesecube MCP Server (`@dcubedllc/cheesecube-mcp`)
MCP server that exposes the Cheesecube Agent API as tools. Use with Claude Desktop, Cursor, or any MCP client.
## Quick start (Claude Desktop)
Add to your Claude Desktop MCP config (e.g. `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS):
```json
{
"mcpServers": {
"cheesecube": {
"command": "npx",
"args": ["-y", "@dcubedllc/cheesecube-mcp"],
"env": {
"CHEESECUBE_API_URL": "https://playform-api-507636286767.us-central1.run.app",
"CHEESECUBE_AGENT_KEY": "your-agent-key"
}
}
}
}
```
`CHEESECUBE_API_URL` defaults to the production API if unset. Write tools require `CHEESECUBE_AGENT_KEY`.
> Until `@dcubedllc/cheesecube-mcp` is on npm, use `@dcubedllc/cheesecube-mcp` as above. After npm publish, switch args to `["-y", "@dcubedllc/cheesecube-mcp"]`.
## Environment
| Variable | Required | Default | Description |
|----------|----------|---------|-------------|
| `CHEESECUBE_API_URL` | No | Production Cloud Run API | Base URL of the Cheesecube API. |
| `CHEESECUBE_AGENT_KEY` | For write tools | — | Agent API key. Required for create/preview/publish cheese and template tools. |
## Tools
| Tool | Auth | Description |
|------|------|-------------|
| `list_engines` | optional | List available engines. |
| `list_templates` | optional | List templates. |
| `get_template` | optional | Get full template by ID. |
| `search_assets` | optional | Search published assets. |
| `create_cheese` | key | Create draft cheese. |
| `preview_cheese` | key | Playable manifest + ingredientUrls. |
| `publish_cheese` | key | Publish a draft cheese. |
| `create_template` | key | Create draft template. |
| `publish_template` | key | Publish a draft template. |
## Local development
```bash
npm install
npm run build
CHEESECUBE_AGENT_KEY=yourkey node dist/index.js
```
## License
MIT
TDQS
Scored across 9 tools
Each tool targets a unique action-resource pair (e.g., list_engines vs. list_templates), and no two tools overlap in functionality. The distinctions are clear and unambiguous.
All tool names follow a consistent verb_noun pattern (list_, get_, search_, create_, preview_, publish_), with no mixed conventions or vague verbs. The naming is uniform and predictable.
Nine tools cover the core operations for managing engines, templates, assets, and cheeses. This is a well-scoped set that is neither sparse nor bloated for the apparent domain.
The set includes create, read (list/get/search), and publish actions, but lacks update and delete operations for both templates and cheeses. These are notable gaps in lifecycle management, though the primary creation and publication flows are covered.