@princesseuh/catalogue-mcp

An MCP server for Erika's catalogue. A list of all the games, movies, TV shows, and books Erika has consumed, with her ratings and reviews.

Install

Claude Desktop

Open the Claude menu > Settings > Developer > Edit Config. Claude Desktop opens claude_desktop_config.json for you (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\). Add:

{
	"mcpServers": {
		"erika-catalogue": {
			"command": "npx",
			"args": ["-y", "@princesseuh/catalogue-mcp"]
		}
	}
}

Restart Claude Desktop to pick up the change.

Claude Code

Run:

claude mcp add --transport stdio erika-catalogue -- npx -y @princesseuh/catalogue-mcp

That writes to ~/.claude.json by default. Add --scope project to write to .mcp.json in the current repo instead.

Codex

Add to ~/.codex/config.toml:

[mcp_servers.erika-catalogue]
command = "npx"
args = ["-y", "@princesseuh/catalogue-mcp"]

opencode

Add to ~/.config/opencode/opencode.json (or a project-level opencode.json):

{
	"$schema": "https://opencode.ai/config.json",
	"mcp": {
		"erika-catalogue": {
			"type": "local",
			"command": ["npx", "-y", "@princesseuh/catalogue-mcp"],
			"enabled": true
		}
	}
}

The first call fetches https://erika.florist/catalogue/mcp.json and caches it to ~/.cache/erika-catalogue-mcp/data.json for 24 hours, unless the --refresh flag is passed.

Environment variables

• CATALOGUE_URL: override the JSON source. For local development: http://localhost:8080/catalogue/mcp.json.

Tools

search_catalogue

Filter the catalogue. All inputs are optional.

| field | type | notes | | ----------------- | --------------------------------------- | ------------------------------------------------------------ | | type | "game" \| "movie" \| "show" \| "book" | restrict to one type | | rating_at_least | 0..5 | 0=Hated, 1=Disliked, 2=Okay, 3=Liked, 4=Loved, 5=Masterpiece | | year | integer | exact original release year | | year_after | integer | released in or after this year (inclusive) | | year_before | integer | released in or before this year (inclusive) | | finished_year | integer | year the media was consumed | | finished_after | YYYY-MM-DD | finished on/after this date | | finished_before | YYYY-MM-DD | finished on/before this date | | genre | string | case-insensitive substring match | | query | string | fuzzy match on title or author, ranked by relevance | | mentions | string | substring match on the review body | | limit | 1–200 (default 50) | |

Returns summaries (id, title, rating, dates, author). Call get_entry for the full review.

get_entry

Returns one full catalogue entry, with the written review as raw Markdown in content.

• type: "game" | "movie" | "show" | "book"
• id: entry slug, e.g. hotline-miami

list_recent

Recently-finished entries, newest first.

• type: optional filter
• limit: 1–100 (default 20)

stats

Counts per type, average rating per type, and the latest finished date.

refresh

Refetch the catalogue from the live endpoint, bypassing the 24h cache. Use this if the catalogue has new entries since the server started.

Examples

// "What games did I like that came out in 2026?"
search_catalogue({ type: "game", rating_at_least: 3, year: 2026 });

// "Show me my recent movie ratings."
list_recent({ type: "movie", limit: 10 });

// "What did I think of Hotline Miami?"
get_entry({ type: "game", id: "hotline-miami" });

Development

pnpm install
pnpm build       # compile to dist/
pnpm dev         # watch mode

Point at a local website build:

CATALOGUE_URL=http://localhost:8080/catalogue/mcp.json node dist/index.js