Stump

Last reviewed Sep 13, 2026 · 768 words

Stump runs in a single container on 256 MB of RAM, serves CBZ, CBR, EPUB, and PDF over OPDS to whatever reader you already own, and has an interface that looks like it was designed this decade. It is also pre-1.0, MIT-licensed, and two people short of the team Komga has. That is the whole trade in one sentence: the nicest lightweight comic server available, with the caveat that "lightweight" and "young" arrive together.

A Rust backend is why it is fast, pre-1.0 is why it is risky

The server is Rust with SQLite underneath; the web client is TypeScript and React, which is why the catalogue lists it under TypeScript while the project describes itself as Rust. On a Raspberry Pi 4 or an old NUC, that shows up as scans that finish while you make coffee and page turns that do not stutter, where a Java server on the same hardware wants 512 MB before it does anything. The 2,692 stars are real interest, and the project has been public since 2022.

The risk is the version number. Pre-1.0 means the database schema can change between releases, and a migration failure on a comic library is annoying rather than catastrophic (the files are untouched; only progress and metadata live in the database). Still, pin the image tag, read the release notes before pulling, and keep the config directory in your backup set.

Running it: one container, one port, one volume

The image is aaronleopold/stump, the port is 10801, and the config and database live in one directory:

services:
  stump:
    image: aaronleopold/stump:latest
    ports:
      - "10801:10801"
    environment:
      - PUID=1000
      - PGID=1000
    volumes:
      - ./config:/config
      - /mnt/library/comics:/comics:ro
    restart: unless-stopped

Mount the library read-only. Stump does not need to write to your files, and a read-only mount means a bug in a young project cannot cost you the collection. On first visit you create the owner account, add a library pointing at /comics, and pick a library pattern: series-based (each folder is a series) or collection-based (the folder tree becomes a hierarchy). Choose once; changing it later means a rescan.

OPDS is how you actually read on a tablet

The web reader is decent for a browser, but the point of a comic server is a native reader with offline caching. Stump exposes an OPDS catalogue, and it supports the page-streaming extension so readers can fetch individual pages instead of a whole 300 MB CBZ. Panels and Chunky on iPad, Moon+ Reader and KOReader elsewhere: add the catalogue URL, which by default is http://your-host:10801/opds/v1.2/catalog, log in with a Stump account, and the library appears.

Reading progress syncs back through OPDS for readers that support it, and through Stump's own API for the official mobile app, which exists but is younger than the server. Treat the mobile app as a preview and OPDS as the production path.

Where Komga and Kavita still win

Three things. Metadata: Komga reads ComicInfo.xml, matches series against external sources, and has a mature editing UI; Stump reads ComicInfo.xml but the editing and matching story is thinner. Ecosystem: Komga has a Mihon extension and years of third-party tooling; Stump has fewer hooks. Breadth: Kavita handles comics, manga, and prose ebooks with per-format readers and a bookmarking and annotation layer Stump does not attempt. If your library is 5,000 comics with fussy metadata and multiple family members who want recommendations, Komga is the safer pick and the ebooks category page covers the rest of the field.

Stump wins on resource use, on the interface, and on features Komga lacks: smart lists with saved filters, and book clubs, where a group reads a series on a schedule with a discussion thread. Those are niche, but they are the reason a small reading group would pick Stump over anything else.

What I'd do

Run Stump if the box is small (a Pi, a 1 GB VPS, a NAS with little headroom) or if the book-club feature is what you came for. Mount the library read-only, pin the image tag, back up ./config nightly, and read from Panels or KOReader over OPDS rather than from the web UI. Run Komga if the library is large, the metadata matters, or other people in the house will complain when something breaks between releases. Either way the files stay yours and in a normal folder, so switching later costs one rescan.