JSON Crack vs Json Web Viewer

JSON Crack made the node-link graph the default way people expect to look at JSON, and it did it as MIT-licensed open source. We build a tool that now ships a graph too. Two things are worth saying before anything else: their graph is better than ours, and you can read their code and not ours.

Where JSON Crack is better

• It is open source, MIT-licensed, with roughly 44,000 GitHub stars. You can audit exactly what it does, run it on your own machine, fork it, or lift pieces of it into your own product. Json Web Viewer is closed source. You can verify our privacy claim at runtime — that is what connect-src 'self' is for — but "verify the behaviour" is a weaker offer than "read the code," and for some readers it is the deciding difference.
• The graph is the product, not one of seven views. Theirs has had far more attention: richer node rendering, layout controls we do not match, and — in the paid ToDiagram tier — the thing our graph fundamentally cannot do, editing through the diagram. Change a node on their canvas and the source updates; edit the source and the diagram redraws. Our graph is read-only: you navigate, collapse, search and export from it, and all editing happens in the text or tree panes.
• Distribution beyond a web page. A Chrome extension and a VS Code extension, so the graph shows up where you already are. We ship a jwv command-line tool and an MCP server instead, which is a different audience, not a substitute.
• Kotlin data classes. Their type generation covers Kotlin; ours covers TypeScript, Go and Rust only.
• Mermaid, custom nodes, and AI (ToDiagram). Their paid tier accepts Mermaid flowchart/sequence/state syntax as an input format, lets you author diagram elements that are not derived from data at all, and drafts diagrams from a plain-language prompt. We do none of those, and the AI one we structurally cannot: an outbound call to a language model is exactly what our Content Security Policy forbids.

The free-tier size ceiling, which cuts the other way

JSON Crack's free editor now publishes a hard limit: approximately 300 KB, varying with data complexity and hardware. That is small — smaller than a single verbose API response in a lot of production systems. Json Web Viewer's documented target is 16 MB, and it is a measured figure: 16 MB is the largest size at which every performance bound we set actually held under test, not the largest that loads without crashing.

ToDiagram removes the ceiling entirely, and its own FAQ says how: "ToDiagram processes diagrams locally in your browser by default. If you explicitly choose server processing from the editor prompt for large diagrams, data is used only to compute layout and is not stored." That is an opt-in with a promise attached, and it is worth knowing that the prompt appears exactly when your document is large enough to be a real production payload rather than a toy. We cannot offer unlimited size for the same reason: there is no server to send it to, and connect-src 'self' would refuse to reach one if there were.

Our own graph has a limit too, and it is a different one: 1,200 visible nodes. Past that, graph mode declines to render and says so, rather than freezing the tab — dagre's layout pass is a single synchronous call that cannot be interrupted once started. Collapse branches or use tree mode for documents that big.

Where Json Web Viewer is better

• It is an editor, not primarily a visualiser. Inline editing in a tree, a sortable editable grid for arrays of objects, a Miller-column drill-down browser, and a full CodeMirror 6 text pane. JSON Crack's centre of gravity is rendering a document you already have.
• A semantic diff. Two panels, compare mode, differences highlighted at the data level so reordered object keys are never reported as changes — plus copying the difference out as an RFC 6902 JSON Patch. We found no diff feature in JSON Crack at all.
• A HAR viewer and HAR diff. Load a browser network capture, read each response body as real JSON rather than an escaped string, and diff two captures to see what changed between them.
• "Across all records." Select a field inside an array of objects and see its value distribution across every sibling record — distinct values, frequencies, how many are null. Nothing equivalent turned up in JSON Crack.
• Genuinely offline. Load the app once and it works with the network disconnected. ToDiagram offers an app download, but its own FAQ states "online access is still required."
• No upsell in the way. JSON Crack now opens with an interstitial recommending ToDiagram and describing its own free editor as "best for small files and quick edits," and keeps an "Upgrade to Pro Editor" link in the toolbar. There is no paid tier here to be nudged toward.

Things that are a wash

Both render JSON as a node-link graph with pan, zoom, fit, collapse, search, and image export (PNG, JPEG and SVG on both sides). Both accept YAML, XML, TOML and CSV as input alongside JSON. Both generate type definitions from a document, generate a JSON Schema from data, generate mock data from a schema, decode a JWT, randomize a document into fixture data, and run jq queries — JSON Crack lists all of these on its own marketing page, and we ship all of them, so this is a genuine draw rather than a claim in either direction. Both are free to use for that whole set.

Which one should you use

Use JSON Crack if the graph is the point, if you want to edit through the diagram (ToDiagram), if you need the browser or VS Code extension, or if being able to read and self-host the source matters to you. All four are good reasons and none of them are things we can match.

Use Json Web Viewer if your document is bigger than a few hundred kilobytes, if you need to edit rather than mainly look, if you want a semantic diff or a HAR viewer, or if you want a privacy guarantee with no prompt that changes the answer when the file gets large.

Other comparisons: JSON Editor Online · json.site · all four side by side. Or try our graph view and judge it yourself.