Calibre-Web

Last reviewed Aug 26, 2026 · 819 words

Calibre-Web will not import your e-books. It reads a metadata.db that Calibre, the desktop program, has already created, and if you point it at a folder of bare EPUB files it will simply refuse to start. Every setup complaint I have seen about this project traces back to that one fact: it is a front end for a Calibre library, not a replacement for Calibre. Accept that, keep Calibre in the loop for ingestion, and what you get is the nicest way to actually read the books you have hoarded.

The library comes first, the container second

Install Calibre on any machine, add your books, and let it build the library folder: one directory per author, one per title, plus metadata.db at the root. That folder is what Calibre-Web mounts. The container itself is small (256 MB of RAM is plenty) and the LinuxServer image is the one most people run:

services:
  calibre-web:
    image: lscr.io/linuxserver/calibre-web:latest
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Europe/London
    volumes:
      - ./config:/config
      - /mnt/books/calibre-library:/books
    ports:
      - "8083:8083"
    restart: unless-stopped

On first visit to port 8083 you log in as admin with password admin123, point the "Location of Calibre database" field at /books, and change that password immediately. The Calibre desktop app stays installed somewhere for adding and fixing books. Calibre-Web can edit metadata and upload files if you enable it, but it does a rougher job than Calibre proper, and the catalogue's "not a metadata manager itself" is the right way to think about it.

Do not let two programs write the same SQLite file

The database is SQLite and Calibre assumes it has exclusive access. If Calibre desktop has the library open over a network share while Calibre-Web writes to it, you will eventually corrupt metadata.db. The safe patterns, in order of preference: run Calibre on the same host, close it before Calibre-Web needs write access, or treat Calibre-Web as read-only (leave uploads and metadata editing disabled) and do all changes in Calibre. If you want a fully web-based pipeline, the LinuxServer team also ships a Calibre container with a web GUI, and there is a popular automated-ingest companion project, but the SQLite rule still applies to every one of them. Back up metadata.db before doing anything clever.

The reader and OPDS are the actual payoff

Two features justify the whole exercise. The first is the in-browser reader, which handles EPUB, PDF, CBR/CBZ comics and a few more formats well enough that a tablet becomes a reading device without any app. The second is the OPDS catalogue at /opds, which lets proper reading apps (KOReader on an e-ink device, Moon+ Reader on Android, several on iOS) browse and download from your library over HTTP with basic auth. For anyone with a Kobo, there is also a Kobo sync endpoint that makes the device pull your library instead of Kobo's store.

Send to Kindle works through an SMTP account you configure in the admin settings: Calibre-Web emails the file to your Kindle address, so you also need to whitelist the sender on Amazon's side. It converts with Calibre's ebook-convert binary if you tell it where that binary lives, which the LinuxServer image can install for you with an optional package mod. Between OPDS and Kindle delivery, this covers the Kindle use case for most people while keeping the files on your disk.

Multi-user is real, and reverse proxy is the usual step after

Each user gets their own shelves, read-state and download permissions, and admins can restrict users by language or tag, which makes a "kids only see kids books" setup a two-minute job. LDAP and OAuth logins exist for households already running an identity provider. Put it behind Caddy or Traefik with HTTPS before exposing it, and be aware that some sub-path setups need the X-Script-Name header; a dedicated subdomain avoids that entirely.

If your collection leans toward comics and manga, Kavita and Komga do that better and do not need Calibre at all. For audiobooks, Audiobookshelf is the right tool and Calibre-Web is the wrong one. The ebooks category has the full field.

What I'd do

Calibre desktop on the same box as the container, library on a path both can see, Calibre-Web read-only with uploads disabled, HTTPS via a subdomain. OPDS turned on for the e-reader, SMTP configured once for Kindle delivery, nightly copy of metadata.db into the normal backup rotation. Add books in Calibre, read them anywhere through Calibre-Web, and never let both write at once. Run that way, an 18,053-star project that started in 2017 has years of boring, reliable service in it.