Online JSON tools compared — honestly, including where we lose

We build one of the tools on this page, so treat everything here with the suspicion that deserves. What we can offer instead of asking for trust is specificity: every claim below names a concrete, checkable capability, every section on a competitor leads with what that competitor does better than we do, and the method and dates behind the research are stated so you can tell how stale it is.

The short version

If you only read one section, read this one. These are real recommendations, not a funnel:

• Your documents are hundreds of megabytes. Use JSON Editor Online. It states a 512 MB ceiling; our own measured, documented target is 16 MB. That is not a close call.
• You want to read the source, self-host it, or fork it. Use JSON Crack. It is MIT-licensed and openly developed (~44k GitHub stars). Json Web Viewer is not open source — you can verify what it does at runtime, but you cannot read its code.
• You live in a spreadsheet-grade grid, or you need Markdown/SQL export, or a non-English interface. Use json.site. Its table does column drag-reorder, column hiding, box-select-and-delete and Cartesian copy; it exports Markdown and SQL; its UI is translated. We ship none of those four.
• You need the document to provably never leave your machine, and you want to check that yourself rather than believe a sentence about it. That is the one axis where this app is not just competitive but structurally different — see the privacy section and the security page.
• You want the widest set of things in one tab — text, tree, table, graph, Miller-column and HAR views, semantic diff, jq, schema generate/validate, type generation, JWT decode — without an account or a paid tier. That is what this app is actually optimised for.

Capability matrix

Rows chosen because they are the ones people actually decide on. "—" means we found no evidence of the capability, which is not the same as proof it doesn't exist; see how this was researched.

Where each competitor is genuinely better

JSON Editor Online

Size, query breadth, and loading a document from a URL. Its stated 512 MB ceiling is 32× our measured 16 MB target. It ships four query languages to our two, so if you specifically want JSONPath or JavaScript-with-Lodash, it has them and we do not. And its Open from URL genuinely works — ours is the honest failure case described below. It also offers cloud-saved documents, which a tool with no server cannot. Full comparison →

JSON Crack (and ToDiagram)

Open source, and a far more mature graph. JSON Crack is MIT-licensed with ~44k stars — you can audit it, self-host it, or fork it, none of which you can do with this app. Its graph is the product rather than one of seven views, and its paid tier, ToDiagram, does something our graph flatly cannot: edit through the diagram, with the source text updating in response. It also ships Mermaid input, custom nodes, a Chrome extension and a VS Code extension. Full comparison →

json.site

The table, the export formats, and the interface itself. Its grid does column drag-reorder, column hiding, box-select-and-delete and Cartesian select/copy — ours does none of those. It exports Markdown and SQL, which we do not support at all. Its keyboard shortcuts are user-rebindable and its UI is translated into multiple languages; ours are neither. It is free, ad-free, and shipping faster than either of the other two. Full comparison →

Where this app is genuinely better

• Breadth in one tab, with no account and no tier. Seven ways to look at the same document — text, tree, table, graph, Miller columns, generated type definitions, and a HAR request browser — plus semantic diff, JSON Patch, jq, JMESPath, schema generate/validate, mock data, JWT decode, and CSV/YAML/XML/TOML/NDJSON/Excel conversion. Every one of those is free, and none of them is behind a sign-in.
• A HAR viewer and diff. Load a browser network capture, read each response body as real JSON, and diff two captures. None of the three competitors do this.
• "Across all records." Select a field inside an array of objects and the node inspector shows that field's value distribution across every sibling record — how many are null, how many distinct values, which value dominates. We found nothing equivalent in the other three.
• Schema generation scoped to one node. Right-click any node in the tree and generate a JSON Schema for just that subtree, rather than only for the whole document.
• The privacy guarantee is enforced, not promised — the section below.

The one axis that isn't a feature

Every tool on this page says its processing happens in your browser. That sentence is worth approximately nothing on its own, because all four of us can write it. What differs is whether anything enforces it:

• JSON Editor Online stores cloud documents on a server, and warns in its own FAQ that public cloud documents are "publicly accessible for anyone who has the document id, so make sure you do not save documents with sensitive information in the cloud." Private documents are the $5/mo tier.
• ToDiagram processes "locally in your browser by default," and its own FAQ says that if you choose server processing for large diagrams, the data is used to compute layout. Opt-in, and they promise not to store it — but the prompt appears exactly when the document is big enough to be a real production payload.
• json.site markets "Private, Secure & Free," keeps history locally, and added a local-data privacy banner in August 2026. It is the closest to our posture. It is also a policy claim with no published enforcement mechanism, and the site is closed-source.
• Json Web Viewer serves connect-src 'self' in its Content Security Policy. The browser refuses any network connection to any other origin — not because our code chooses not to, but because the browser blocks it. There is no prompt, no upgrade path, and no large-document escape hatch that changes that answer. Open DevTools' Network tab and check the response headers yourself; that takes about a minute and requires believing nothing on this page.

The honest cost of that stance is on this page too, and it is not small: no cloud documents, no collaboration, no AI features, and no working "open from URL." Those are not things we haven't got around to — they are things this architecture forbids. In November 2025, security researchers reported finding tens of thousands of saved entries exposed on other online JSON tools, containing cloud keys and credentials. That is the failure mode the trade-off buys out of.

How this was researched, and how stale it is

Competitor capabilities come from hands-on examination of the live products on 2026-07-31, re-verified against the live sites and their own changelogs and pricing pages on 2026-08-12. Where a fact came from a competitor's own changelog or marketing rather than from driving the product, the tables above say "per its own marketing" or "unconfirmed" rather than presenting it as observed. json.site in particular ships every few days; assume it has moved since.

Claims about this app were checked against the shipped build, which is why some of them are less flattering than our own marketing elsewhere. If you find something on this page that is wrong — especially something wrong in our favour — tell us at contact@jsonwebviewer.com and we will correct it.

Detailed comparisons: JSON Editor Online · JSON Crack · json.site. Or open the editor, browse every tool, or read how the privacy claim is enforced.