This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
A small MCP server for Grimmory, the self-hosted ebook, comic and audiobook library. It lets an AI assistant such as Claude Code search your library and tidy metadata, tags, genres and shelves.
Built and tested against Grimmory v3.5.0. Grimmory marks its API as unstable, so check it again after you upgrade.
Tool | Kind | What it does |
| read | Free-text search plus facet filters ( |
| read | Lists the authors, genres, tags, series, publishers and shelves already in use, so new values match existing ones |
| read | One book's metadata, shelves and locked fields |
| read | Regular shelves visible to the user |
| write | Changes chosen fields on one book and returns before/after values |
| write | Tags, genres, moods, authors, series, publisher or language on up to 100 books; shows a preview first, writes only on |
| write | Creates a shelf; refuses duplicate names |
| write | Adds books to shelves or removes them |
Related MCP server: Calibre MCP Server
These rules are enforced in code, not in a prompt:
No delete tools. Nothing can delete books, files, shelves, users or metadata values.
Locks are respected. A field locked in Grimmory is refused, never overwritten. No tool can change locks.
No accidental wipes. Grimmory's metadata endpoint defaults to replaceMode=REPLACE_ALL, which clears every field you don't send. This server always sends REPLACE_WHEN_PROVIDED.
Bulk edits are two-step. The first call returns a per-book preview. Nothing is written until the same call is repeated with confirm=True.
Nothing is cleared. Empty values are refused. Remove the last tag from a book in the Grimmory UI.
Use a dedicated Grimmory user for the assistant. The server is the only thing between the assistant and whatever that user is allowed to do.
Requires Python 3.12+ and uv.
git clone https://github.com/jensopt/grimmory-mcp
cd grimmory-mcp
uv sync --frozenVariable | Required | Meaning |
| yes | Base URL, e.g. |
| yes | Grimmory username (not email; Grimmory logs in by username) |
| yes | Its password |
| no | Only if Grimmory sits behind Pangolin, see below |
| no | Second half of the same token |
claude mcp add grimmory -s user \
-e GRIMMORY_URL=https://books.example.com \
-e GRIMMORY_USERNAME=assistant -e GRIMMORY_PASSWORD=... \
-- uv run --quiet --frozen --project /path/to/grimmory-mcp grimmory-mcpRestart Claude Code afterwards, because MCP servers only load at startup.
Pangolin's header authentication won't work for this server. It is HTTP Basic auth in the Authorization header, and Grimmory needs that same header for its Bearer token. A request can only carry one.
Use a shareable link (access token) on the Grimmory resource instead. Pangolin shows it as ?p_token=<id>.<token>. Put the part before the dot in PANGOLIN_TOKEN_ID and the part after it in PANGOLIN_TOKEN. The server sends them as P-Access-Token-Id / P-Access-Token headers, which leaves Authorization free for Grimmory.
Access tokens last 2 hours. On v3.5.0, /api/v1/auth/refresh returned 401 in testing, so the server logs in again when the token expires.
Grimmory calls genres categories on a single book and genres in bulk edits. The tools use the API's names.
To see the full API, set API_DOCS_ENABLED=true on your instance and read /api/openapi.json.
Uses the mcp Python SDK 2.x (MCPServer, formerly FastMCP).
MIT
Read-only access to Penwright's writing-craft book library, grounded guide metadata (card-level...
The media memory layer for AI agents and their humans. Your AI client can search your collection, add and log what you've finished, update ratings and lists, and find patterns across everything you've read, watched, played, and listened to.
Academic literature search, retrieval, and private library management on top of OpenAlex.
Manage your Readlink book library: add books, track reading, rate, notes, highlights, shelves.
