A collapsible tree is a correct rendering of JSON. It is not always a useful one. 🌳
If the payload is an API response, you usually want one of two things: relationships (this object points at that list) or records (this array should be columns). BracketView already had Monaco text, a virtualized tree, and a graph. This release adds the two shapes that were missing, plus a few tools that used to live in another tab.
- 🌐 Web: https://app.bracketview.in
- 💻 Windows: https://bracketview.in/downloads
Node and table views are free. You do not need Pro to click them. ✅
🔗 Node view
Node view is a canvas of linked cards, not a force-directed hairball of every key.
- Nested objects become their own nodes.
- Arrays of objects render as tables on the canvas.
- Edges carry the parent → child relationship.
- Search focuses a matching node instead of expanding the whole tree.
- Minimap is optional so the canvas stays readable.
Implementation notes, for people who care: React Flow + dagre for layout. There is a hard cap (MAX_FLOW_NODES = 250) so a pathological document does not freeze the tab. Tables on the canvas also cap columns and rows. If your file is huge, stay on tree / table / text (Performance Mode on Pro is for that path). Node view is for the payload you want to inspect as a diagram.
📊 Table view
Table view is for arrays of objects sitting inside a larger document.
You do not have to flatten the JSON into CSV first. Nested values expand in the cell. Arrays longer than a threshold collapse until you open them. Rows virtualize after ~80 so a large list is still scrollable. Primitives get type coloring consistent with the tree.
0.1.3 is a polish release on this view: long strings wrap and clamp instead of overlapping the next column (table-layout: fixed, line-clamp, wrap-anywhere). That sounds boring until you paste a real webhook body with a JWT in a field. 😅
🔐 Encoder / decoder and text compare
Under More:
| Tool | What it does |
|---|---|
| 🔐 Encoder / decoder | Base64, URL, ASCII, Unicode, binary, hex, Base64 URL-safe, ROT13, Morse |
| ↔️ Text compare | Monaco DiffEditor, side by side from md and up |
These run in the browser. In the Windows app they stay on the device.
- Encoder: https://app.bracketview.in/encoder-decoder
- Compare: https://app.bracketview.in/compare-view
Structural JSON diff is still the separate Diff tool. Text compare is for logs, headers, and “why do these two strings disagree.”
⚙️🌍 Workspace settings and nine locales
The header gear is Workspace settings: appearance, editor font, and language. Font and theme are no longer buried on the profile page.
Locales: en, zh-CN, ja, ko, pt-BR, es, fr, de, ru.
💻 Windows desktop (Tauri 2)
The desktop shell is a local JSON workspace: viewer (text, tree, graph, node, table), jq/JSONPath, stats, encoder, text compare, native Open / Save. Cloud, auth, and marketing stay out of that binary.
Facts only:
- 🪟 Windows 10 / 11, NSIS + MSI
- 🐧 Linux: not shipped
- 🍎 macOS: no build
- ⚠️ Installers may trip SmartScreen until they are Authenticode-signed; More info → Run anyway
- Settings can check latest.json and send you to the installer
Shortcuts that only exist on desktop: Ctrl+O open, Ctrl+S save, Ctrl+Shift+S save as. Share-snapshot shortcut stays web-only.
🌐 Web-only UX in the same drop
If you use the hosted app:
- 👤 Profile photo opens Manage Profile / Sign out (header Sign out button is gone)
- 🔍 Tree search sits on the tree toolbar; ✨ AI slides in from the right, next to keyboard shortcuts
- 🕒 Profile JSON history: custom date filter + copy JSON in one click
- ⚡ Route changes show a layout skeleton instead of a blank page
🔒 Privacy model (unchanged)
| Local (default) | Optional server |
|---|---|
| View, format, validate, JSONPath, jq, client-side diff/schema/types, encoder, text compare | AI (server-side Groq/Gemini keys), encrypted snapshots, webhook storage, signed-in history |
Analytics may record events without collecting raw JSON. Details: bracketview.in/privacy.
BracketView is still not an HTTP client. Paste from your client, or use Webhook Tester for inbound events.
🚀 Try it with a real payload
- Open app.bracketview.in
- Paste something you actually received this week
- Cycle Tree → Graph → Node → Table
- If you want it offline, grab the Windows build
If node view truncates your document or table view picks the wrong columns, that is useful feedback — those caps and heuristics are the interesting part, not the screenshots. 🙏
Top comments (2)
Capping Node view at 250 nodes is a sensible product decision; a readable partial diagram is more useful than a React Flow canvas that freezes on a pathological payload. The table view's fixed layout, line clamping, and wrap-anywhere handling also targets the kind of failure that only appears with real webhook bodies, especially JWT-sized fields. I'd make truncation and column-selection heuristics visibly explain themselves in the UI, because once a tool transforms data for readability, users need to know whether they are seeing the whole payload or an intentional projection of it.
Thanks Marcus — that’s exactly the tradeoff. The 250-node cap is there so a pathological payload doesn’t freeze the canvas; a readable slice beats an unusable full graph.
You’re right that once we project, we should say so. Node view already shows “Showing n of m nodes” when we hit the cap. Table view is weaker: long strings clamp (full value on hover) and columns are the union of keys, without a visible “this is a projection” cue. I’ll make truncation and column heuristics first-class in the UI so it’s obvious whether you’re looking at the whole document or a readability view.