Komga

Last reviewed Sep 13, 2026 · 739 words

Komga reads a folder as a series and each file inside it as a book, so how you organise the disk decides how good the server is before you install anything. Get the layout right and Komga is a 5-minute deployment that turns a pile of CBZ files into a library any comic app can browse. Get it wrong and you spend a weekend merging duplicate series. Sort the files first; the rest of this is easy.

Folder equals series, file equals book

The expected layout is one directory per series with the issues or volumes inside it, named so they sort correctly:

/comics/
  Saga/
    Saga 001.cbz
    Saga 002.cbz
  Berserk/
    Berserk v01.cbz

Supported formats are CBZ, CBR, PDF and EPUB. If an archive contains a ComicInfo.xml, Komga reads the title, number, writer, summary and series metadata from it, which is what most tagging tools write. Loose one-shots that do not belong to a series go in a directory whose name matches the oneshots setting, so they do not each become a one-book series. Nested folders inside a series are scanned but easily produce surprises, so keep it flat.

The compose

services:
  komga:
    image: gotson/komga:latest
    ports:
      - "25600:25600"
    volumes:
      - ./komga-config:/config
      - /mnt/media/comics:/data:ro
    environment:
      - TZ=Europe/London
    user: "1000:1000"
    restart: unless-stopped

Open port 25600, create the admin account on the first visit, add a library pointing at /data, and wait for the scan. The media mount is read-only on purpose: Komga never needs to write to your archives, and a typo in a metadata edit should not be able to touch them. It is a JVM application, so the 512 MB in the listing is real rather than nominal, and a library of tens of thousands of files is happier with 1 GB.

OPDS and the reader apps are the point

The web reader is good, with a webtoon mode for vertical strips and page-spread handling for double pages, but the reason people run Komga is what connects to it. It exposes OPDS feeds, which means any OPDS-capable app can browse and download from it: Mihon on Android through its Komga extension (with read progress synced back), Paperback and Panels on iOS, Chunky on iPad, KOReader on e-ink devices. The REST API behind those is clean and documented, which is why third-party tools target Komga first. If your reading happens on a phone or tablet rather than in a browser, this is the deciding feature.

Metadata is good, and Komf makes it better

Komga's own metadata handling is the strongest among self-hosted comic servers: per-series and per-book fields, collections, read lists, reading direction, and edits that survive rescans. What it does not do is fetch metadata from the internet. Komf fills that gap as a companion service that scrapes sources such as ComicVine, MangaUpdates and AniList and writes the results into Komga through the API. For a manga-heavy library it is worth running from the start.

Where Kavita wins

The listing's "less ideal for plain e-books" line is fair. Komga opens EPUBs, but its reader treats a book as a sequence of pages, which suits fixed-layout comics and suits a 400-page novel badly. Kavita has a genuinely good reflowable EPUB reader, reading lists that mix formats, and a more Netflix-like browsing experience, at the cost of slightly less careful comic metadata. If your library is mostly comics and manga, Komga. If it is mostly novels, Kavita or Calibre-Web. If it is both, I run Komga for the comics and one of the others for the books rather than forcing either to do the job it is weaker at; the ebooks category has the rest of the field.

What I'd do

Spend the first hour on the disk, not the server: one folder per series, consistent numbering, one-shots in their own directory. Then the compose above with the media mount read-only, Komf alongside it if you collect manga, and Mihon or Paperback on the devices you actually read on. For a collector leaving ComiXology with a few thousand issues, that is a complete replacement on a Raspberry Pi 4 or any old laptop, and it will never delete, re-encode or rename a file you spent years collecting.