Omni Tools

Last reviewed Aug 26, 2026 · 764 words

Omni Tools is a static web app in a container. The server does nothing except hand your browser a bundle of JavaScript, and every conversion, crop, hash and regex then runs inside the tab. That is the entire pitch, and it is a good one: the API response you pasted, the customer CSV, the screenshot that needed resizing, none of it touches a third-party server, and the container idles well under the 128 MB the catalogue lists as its floor. I run it daily. I also still keep CyberChef bookmarked, and the rest of this guide explains why both of those are true.

Deployment is a port mapping and nothing else

There is no database, no volume, no environment variable, no account. The whole install is this:

services:
  omni-tools:
    image: iib0011/omni-tools:latest
    ports:
      - "8080:80"
    restart: unless-stopped

Open port 8080 and you have the searchable tool list. Put it behind Caddy if you want a hostname; there is nothing to authenticate because there is nothing stored. Updating is docker compose pull && docker compose up -d, and since the app carries no state, a bad update rolls back by pinning the previous tag. The MIT licence means you can also just build it yourself if the image ever goes stale.

"Runs in your browser" is literal, which cuts both ways

The tool families are image (crop, resize, convert, compress), video and audio (trim, convert, extract), PDF (merge, split, rotate), plus text, list, CSV, JSON, number and time utilities. The media tools work because FFmpeg has been compiled to WebAssembly and ships inside the page. The upside is privacy and a server that never breaks a sweat. The downside is that the computer doing the work is the one running the browser: a 1.5 GB video trim will pin a laptop core for a few minutes and will simply fail on a phone with 4 GB of RAM. Size your expectations by the client, not the host. For anything server-side and heavy on PDFs, Stirling PDF does OCR and batch jobs that a browser tab should not attempt.

The catalogue's "tool depth varies" is fair

The project's first release was in 2024, and 10,087 stars in that time says the demand was real. It also means the breadth arrived faster than the depth. The JSON formatter, list sorters and image converters are solid; some tools are a single input box with one option, where the equivalent in IT-Tools has 6. Speaking of which: IT-Tools is the older, developer-leaning collection (JWT decoding, UUIDs, hashes, crontab, Docker run to compose). The overlap is large but the emphasis differs. Omni Tools leans towards media and everyday files; IT-Tools leans towards things a backend developer needs at 4pm. The two together use less than 300 MB, so I run both on the same box and search whichever fits the job.

Where CyberChef still wins

CyberChef is a pipeline. You chain 6 operations (from base64, gunzip, parse JSON, extract field, hash, to hex), watch the output update live, and save the whole recipe as a URL you can hand to a colleague. Omni Tools is one tool at a time, with no way to feed the output of one into the next except copy and paste. For decoding malware strings, reversing an odd encoding, or any forensic work, CyberChef remains the tool. For "compress these 40 PNGs to under 200 KB each" or "merge these 3 PDFs", Omni Tools gets there in fewer clicks.

Do not bother exposing it

Because there is no login and no data, exposing Omni Tools to the internet is harmless in the strict sense, and pointless in the practical one: the whole value is that it runs on hardware you control, for people you know. Keep it on the LAN or behind Tailscale, give it a short hostname, and put a tile for it on your dashboard next to the rest of the dev tools.

What I'd do

Deploy Omni Tools and IT-Tools side by side on any box that already runs Docker, both behind the reverse proxy you already have, both LAN-only. Bookmark the local Omni Tools URL in the browser bar where the online converters used to be, and keep CyberChef for the once-a-month puzzle. Total cost is about 250 MB of RAM and 5 minutes, and the online tool sites can stop learning what your data looks like.