One binary
No JVM, no runtime: the whole server is one Rust executable for Linux, macOS, and Windows. Docker image included.
KMServer
Your library. One server.
A full-featured Komga drop-in replacement in Rust: the same formats, API and database, plus additional capabilities — in one self-contained binary with no JVM and a small memory footprint.
Written in RustKomga-compatibleKomga + enhancementsCBZ · CBR · PDF · EPUBExisting librariesExisting databaseCross-platformSingle binarySmall footprint
terminalone command
docker run -d \
--name=kmrs \
--user 1000:1000 \
-p 25600:25600 \
--mount type=bind,source=/path/to/config,target=/config \
--mount type=bind,source=/path/to/data,target=/data \
--restart unless-stopped \
ghcr.io/kmworks/kmrsTry it in your browser.
A live KMServer with the bundled web UI and a public-domain library of English and Japanese comics and ebooks. Everything resets daily.
Email
demo@kmworks.datePassword
kmworks-demoWhy KMServer?
The Komga experience, rebuilt in Rust: one self-contained binary instead of a JVM runtime, and a footprint that stays small.
Small footprint, efficient serving
A 16 MiB SQLite page-cache budget per database, capped allocator arenas, and bounded container caches keep memory predictable and page serving efficient. Enhancements
Web UI included
The kmweb UI is embedded in the binary, so your library works in a browser out of the box.
Komga-compatible, verified
Same port, same mounts, same environment variables, same API.
API parity, verified
REST, OPDS v1.2/v2, SSE, Kobo and KOReader sync. A differential harness compares about 105 endpoints against a live Java Komga instance.
python3 tests/diff/diff.py \
--java-jar komga.jar \
--rust-bin ./kmrsDrop-in replacement
Same port (25600), same /config and /data mounts, same KOMGA_* environment variables. Your existing setup carries over.
Keep your library. Keep your database.
Point KMServer at an existing Komga data directory and it upgrades database.sqlite in place with byte-for-byte Flyway migrations. The Java version can still open libraries written by KMServer, so you can switch back at any time.
Same library
Your files stay where they are.
Same database
Upgraded in place, readable by both servers.
Same API
Your clients reconnect without changes.
Up and reading in one command
Install
One docker run, or a single binary from the releases page.
Keep your library
Point /config at your existing Komga data. KMServer upgrades it in place.
Pick a reader
The bundled web UI, KMReader, KOReader, Kobo, or any OPDS client.
docker run -d \
--name=kmrs \
--user 1000:1000 \
-p 25600:25600 \
--mount type=bind,source=/path/to/config,target=/config \
--mount type=bind,source=/path/to/data,target=/data \
--restart unless-stopped \
ghcr.io/kmworks/kmrsservices:
kmrs:
image: ghcr.io/kmworks/kmrs:latest
container_name: kmrs
user: "1000:1000"
ports:
- "25600:25600"
volumes:
- ./config:/config
- ./data:/data
restart: unless-stopped# grab the archive for your platform from the releases page:
# https://github.com/kmworks/kmrs/releases/latest
./kmrs # serves on http://localhost:25600Beyond compatibility
Capabilities the Java version does not have.
CJK cross-search
Simplified and traditional Chinese match each other, and CJK unigrams find titles mid-run. Search
简繁Smart lists
Save any search as a live list: the filter re-evaluates for whoever is looking, with share scopes and generated mosaic covers. Smart lists
Reading stats
Per-user totals, day-by-day activity, and reading streaks, computed on the server. Reading stats
WebhooksHistory eventsEPUB 2 metadataNatural sortkomf integrationReading trackersFile-backed thumbnailsNext/previousServer stats
The full lists live in the docs: Enhancements, Compatibility, Known limitations, Development.
Bring your own reader
Compatible with the Komga API and its reader ecosystem.