Desktop app · free and open source
This site's query engine as a native app. Open JSON files far bigger than a browser tab can hold, ask them questions in jq, JSONPath, JMESPath or JSON Pointer, and merge, diff, patch and validate documents. Everything happens on your computer, and nothing needs a network.
Get the appSee the source on GitHub
The site is for a quick question about a document you can paste. The app is for the files that do not fit in a tab, and for people who would rather keep their data off the network altogether.
A file of 256 MiB or more is not loaded into memory: it is mapped, indexed once and read as you look at it, so a gigabyte of JSON takes megabytes, not ten times its size. Queries stream, first(…) and limit(…) really stop early, and a new query cancels the one before it.
Your files are read from your disk and never uploaded. The app has no accounts, no analytics or telemetry, no advertising and no update check of its own. The one request it ever makes is for a web address you type into its Source field.
jq (with @csv, @tsv, IN, JOIN and tostream added), JSON Pointer, JSONPath (RFC 9535) and JMESPath. Pick one or let the app tell from what you type.
Snowflake IDs and Postgres bigints come out of a query byte for byte, instead of being rounded through a floating-point number on the way.
Each step of a query is tinted in its own colour, in the same palette as the tutorial, so a long pipeline reads at a glance. Results stream into a tree you can fold or into plain text you can copy. Press Ctrl+F to find, list every match with Find All, or right-click a result to see where in the source it came from.
Type . or [ and the app offers the keys and positions of the document you loaded, with a sample value beside each. Inside select( it offers the fields of each record. What you accept is spelled for the language you are in, so .["first name"] in jq and $["first name"] in JSONPath.
Open it from the toolbar, or press F1. Each of the four languages has a tree of short lessons, every example is taken apart piece by piece with its live result, and Try it loads the example's data and query into the main window so you can change them. The same lessons are on this site, in the Learn section.
The query, the source and the results can each open in a window of their own: a long query on a second monitor, or the results beside a source that has the whole height. Closing the window, or pressing its dock button, puts the pane back.
The toolbar's Tools window has a tab for each. They share one layout: what goes in on the left, what comes out on the right, and a status bar that says what happened. Each one is also on this site, in your browser, for documents of up to a few megabytes.
Combine several files into one document with a jq filter, the way jq -s does. Ready-made filters append, de-duplicate, deep-merge or bundle by file name; edit one into your own.
Pretty-print with 2 or 4 spaces or a tab, minify, sort keys at every depth, escape what is not ASCII. Numbers come out exactly as they went in, however many digits.
Two documents side by side like a file-comparison tool, what differs marked, every change listed with its JSON Pointer, and the RFC 6902 patch that turns one into the other. Move a difference into the other document with a click on its arrow, or only the lines you pick; double-click a line of either side and type over it; and save either document.
Apply an RFC 6902 JSON Patch or an RFC 7386 merge patch. A patch is all or nothing, and when it fails the error names the operation.
Check a document against a JSON Schema (drafts 4, 6, 7, 2019-09 and 2020-12) and list every problem with its pointer and what is wrong. Nothing is ever fetched.
| What | This site | Desktop app |
|---|---|---|
| Where it runs | In a browser tab, on the same engine | On your computer; needs no network |
| Size of a file | Up to 512 MB (the box edits up to 10 MB) | Any size; from 256 MiB it is read from disk, not loaded |
| Query languages | jq, JSONPath, JMESPath, JSON Pointer | The same four, with autocomplete |
| Merge, diff, patch, validate | Yes, for documents of a few megabytes | Yes, for up to 128 MB at once (a setting) |
| Getting JSON in | Paste, drop or pick a file | Also a web address or a path, several files at once |
| Keeping things | Nothing is kept: copy or download what you need | Files on your disk, where you choose |
| Setting up | Nothing | One command, or a download |
Free, with the source on GitHub under the MIT licence. Pick what suits your system; the commands copy with one click.
brew tap nujufas/jsonquery-gui
brew install jsonquery-guiscoop bucket add jsonquery-gui https://github.com/nujufas/scoop-jsonquery-gui
scoop install jsonquery-gui/jsonquery-guiThe program is not code-signed, so Windows SmartScreen may warn the first time it runs.
Each release has an AppImage and a tar.gz for Linux (x86-64 and ARM), and a zip for Windows. The AppImage adds itself to your menu the first time it runs.
macOS builds are not published yet; the app builds from source with cargo run --release -p jsonquery_gui.
On a native Wayland desktop a file cannot be dropped on the window, a gap in the windowing library; the file picker, pasting and the Source field all work, and the README shows the one-click way to start it under X11.