MAL MCP Server

An MCP server for MyAnimeList. It works with any MCP-compatible client or agent (Claude Desktop/Code, Cursor, VS Code, Cline, Continue, and others), since the server speaks the standard MCP stdio protocol.

What it does

It uses a hybrid backend:

• Reads → Tenrai (free, unofficial MAL API). Search, details, rankings, seasons, characters, recommendations and reviews. No credentials required.
• Your personal list → official MyAnimeList API. Read, update and delete entries on your own anime/manga list. Requires a one-time login (see below).

What you need (and what it gets you)

Nothing is required to get started. You can skip straight to Install. Everything below is optional, and each step just adds more:

Without a login, the personal-list tools reply with a clear message telling you how to get one. Everything else keeps working regardless.

Example queries

Once it's connected, just ask your agent in natural language.

No credentials needed (search, details, rankings, seasons, characters, …):

"Search for the anime Frieren and show its score, studio and synopsis."
"What are the top 10 anime of all time?"
"What's airing this season? Sort by popularity."
"Recommend anime similar to Steins;Gate."
"What are people saying in reviews of Chainsaw Man?"
"Show the main characters and Japanese voice actors of Cowboy Bebop."
"When does the next episode of One Piece air?"
"Give me the top manga in the Romance genre."
"What's on the upcoming season's schedule?"
"Pick a random highly-rated anime for me."

With your MyAnimeList account (after a one-time login_mal; see Connect your MyAnimeList account):

"Show my MAL profile and watching stats."
"What's on my anime list that I marked as watching?"
"Add Frieren to my plan-to-watch list."
"Set Cowboy Bebop to completed with a score of 9."
"Bump my episode count for One Piece to 1095."
"Remove Bleach from my manga list."

Tools

token = needs a MyAnimeList login (run login_mal once). Prompts: recommend_similar, seasonal_overview, hidden_gems. recommend_similar's title is optional: omit it and it'll ask which anime you mean. Clients that support prompt-argument completion get live title suggestions as you type.

Install

As an .mcpb bundle (one-click, e.g. Claude Desktop)

Download mal-mcp.mcpb (always the latest release) and open it with your MCP client. It prompts for an optional MyAnimeList Client ID. Set it and run the login_mal tool to enable the personal-list tools (see Connect your MyAnimeList account). Leave it blank to use just the credential-free read tools.

From source

git clone https://github.com/Grinv/mal-mcp
cd mal-mcp
npm ci
npm run build

This produces a self-contained dist/index.js. Point your client at it (see below).

Connect it to an MCP client

Add the server to your client's MCP config (Claude Desktop/Code, Cursor, VS Code, Cline, …). The simplest option is npx, which needs no clone and no build:

{
  "mcpServers": {
    "mal": {
      "command": "npx",
      "args": ["-y", "mal-mcp"],
      "env": {
        "MAL_CLIENT_ID": "..."
      }
    }
  }
}

Or with the Claude Code CLI:

claude mcp add mal -e MAL_CLIENT_ID=... -- npx -y mal-mcp

If you built from source instead, replace "command": "npx", "args": ["-y", "mal-mcp"] with "command": "node", "args": ["/absolute/path/to/mal-mcp/dist/index.js"].

The env block is optional. Omit it to use only the credential-free read tools (search, details, rankings, …); the personal-list tools will return a clear error until you log in. To enable them, set MAL_CLIENT_ID and run the login_mal tool once (a one-time browser authorization; the token is then stored and refreshed automatically). The server does not read a .env file, so pass config via this env block (or your shell environment). See docs/auth.md for the full login walkthrough and docs/clients.md for more clients.

Connect your MyAnimeList account (for the personal-list tools)

The search/browse tools work with no setup. To use the personal-list tools (get_my_*, update_my_*, delete_my_*), authorize your account once. It takes about two minutes. There is no client secret: mal-mcp uses the modern public-client flow (PKCE), so you only need a Client ID.

Step 1: Register a MyAnimeList app (one minute).

1. Go to https://myanimelist.net/apiconfig and click Create ID.
2. App Type: choose other. (Not web: that type forces a client secret this server doesn't use.)
3. App Redirect URL: enter exactly
   Code

   Copy

   http://localhost:8080/callback
   

   (If port 8080 is already used on your machine, pick another port here and set MAL_OAUTH_PORT to the same number. See Configuration.)
4. Fill the remaining required fields with anything reasonable, accept the terms, and Submit.
5. Open your new app and copy its Client ID.

Step 2: Give the Client ID to the server.

Add it to your MCP client config (or paste it into the Claude Desktop install form). Nothing else is needed here:

"env": { "MAL_CLIENT_ID": "paste-your-client-id-here" }

Restart the server/client so it picks up the value.

Step 3: Log in (one click).

In your assistant, run the login_mal tool (or just say "log in to MyAnimeList"). It replies with a link. Open the link, sign in to MAL, and click Allow.

• Running locally (Claude Desktop, or Claude Code on your own machine): login finishes automatically the moment you click Allow. Confirm with "show my MAL profile" (get_my_user_info).
• Running on a remote/SSH/headless host: after clicking Allow, your browser lands on a page that won't load. That's expected. Copy the full address from the browser's address bar (it contains ?code=…) and give it to the submit_mal_redirect tool to finish.

That's it. The token is saved locally (~/.config/mal-mcp/tokens.json, 0600) and refreshed automatically from now on, so you won't need to log in again.

Prefer no interactive step? You can instead pre-set MAL_REFRESH_TOKEN (with MAL_CLIENT_ID) or a standalone MAL_ACCESS_TOKEN. See docs/auth.md for how to obtain them by hand.

Configuration

All configuration is via environment variables, all optional. Without credentials the read tools still work. For the personal-list tools, set MAL_CLIENT_ID and run the login_mal tool once; the access token is then fetched and refreshed automatically. (mal-mcp is a public PKCE client, so there is no client secret.)

Tuning (rarely needed)

These have sensible defaults; change them only if you need different timing, a different upstream base URL, or want to override where/how the server logs or stores its token.

Provide these in your MCP client config's env block (the server does not read a .env file). See docs/auth.md for how to obtain the credentials, and docs/clients.md for client configuration snippets.

NSFW content

NSFW (adult) results are not filtered by default. The server returns whatever the upstream API provides. Most read tools accept two filter levels:

• sfw: true: excludes adult/explicit-rated entries (R+ Mild Nudity and up).
• sfw_strict: true: also excludes anything tagged with the Ecchi genre, even otherwise mainstream, safely-rated shows that sfw alone still lets through (e.g. No Game No Life, Kill la Kill).

On a list/search tool (search_anime, search_manga, get_seasonal_anime, get_upcoming_season, get_anime_schedule, get_top_anime, get_top_manga, get_random_anime, get_random_manga, get_recent_anime_recommendations, get_recent_manga_recommendations) this filters which entries come back. On a by-id lookup (get_anime, get_manga, get_anime_recommendations, get_manga_recommendations, get_character, get_person, get_anime_news, get_manga_news) the item you asked for is always returned regardless. It instead filters NSFW entries out of that item's own nested lists (e.g. an anime's relations, a person's own anime/manga credits), except on the two news tools, where it can empty the result entirely for an NSFW-rated anime/manga. get_top_people/get_top_characters have no such filter, since people/characters aren't independently content-rated the way anime/manga are.

search_anime and get_top_anime/get_seasonal_anime/get_upcoming_season also accept a rating filter for finer-grained control: target one or more specific MAL content ratings (e.g. pg13, r17) directly instead of a blanket adult-content cutoff.

Development

npm run build        # type-check + bundle to dist/
npm test             # node:test suite (mocked, offline)
npm run test:coverage
npm run lint
npm run format
npm run check:api    # live health-check of upstream endpoints
npm run inspector    # run under the MCP Inspector

Runtime requires Node ≥ 20.11 (global fetch, AbortSignal.any). See AGENTS.md for contributor/agent guidance.

Updating

To be notified of new versions, click Watch → Releases on GitHub.

• .mcpb bundle: download the new mal-mcp.mcpb from the releases page and reinstall it in your client (it replaces the old version).
• From source: git pull && npm ci && npm run build.
• npx (after npm publish): unpinned npx -y mal-mcp fetches the latest on the next run; if you pinned a version, bump it.

See the CHANGELOG for what changed in each release.

Attribution and terms

This is an unofficial project and is not affiliated with or endorsed by MyAnimeList. Read data is provided by Tenrai, a free, unofficial MAL API; please respect its rate limits. Personal-list operations use the official MyAnimeList API under your own account and token. Use is subject to the MyAnimeList Terms of Service.

Privacy Policy

mal-mcp runs locally on your machine and has no telemetry of its own. See PRIVACY.md for exactly what data it handles (including what the personal-list tools can read and change on your real MyAnimeList account), what it sends to Tenrai/MyAnimeList, and what (if anything) it stores on disk.

Security

See SECURITY.md for the read/write tool distinction, token storage, credential redaction, and how to report a vulnerability.

License

MIT © Grinv