JSON viewer comparison: JSONBolt vs Dadroit vs JSON Crack vs JSON Hero.

Every JSON viewer claims to be fast until you give it a real file. This is an opinionated, honest JSON viewer comparison from the team that ships one of the four — useful even if you end up picking a different tool.

Disclosure

We make JSONBolt. We use the other three tools regularly. Where we cite specific JSONBolt numbers in this post, they're the published figures on the homepage benchmark (M2 Pro / 32 GB, macOS, cold cache). Competitor timings are representative ranges based on hands-on testing, not head-to-head benchmarks we can replicate-on-demand — treat them as directional, not authoritative. We've tried to be honest about what each tool actually does; if a behavior described here doesn't match the current build of a tool, email us and we'll update.

The competitors in this best JSON viewer comparison: JSON Crack, JSON Hero, and Dadroit JSON Viewer. All three are good tools. We'll say so where they're good. We'll also say where they aren't.

The four contenders at a glance

Test 1: opening a 100 MB JSON file

We used a 100 MB dump of public NYC taxi trip records — flat-ish, ~480 K objects, a few levels deep. This is the file where most viewers separate themselves from each other.

JSON Crack accepts the file — it advertises support up to ~512 MB. Loading is slow and the node-link graph is dense to the point of being a wall of dots at this size; you'd want to switch to its tree/text view to actually read anything. Working as designed, but the visualisation is doing more harm than good here.

JSON Hero is intended for smaller API responses; in our hands a 100 MB upload was outside its practical envelope. Past historical issues at ~22 MB on the hosted version suggest you'd want to keep it under a couple of tens of MB.

Dadroit opens the file to a usable tree quickly — under a few seconds in our hands-on use, with smooth scrolling on the top levels. Genuinely a respectable native-app result; we wouldn't be surprised to see the same on your hardware.

JSONBolt first paint: ~50 ms (the published number on the homepage, which matches our local runs). The whole file isn't parsed at first paint — it's mmap'd, the structural pass is running on a background thread, and the tape fills in as you scroll. If you want the gory detail of how this works, see how we parse JSON at 2 GB/s.

Test 2: opening a 2 GB JSON file

Concatenated event-log dumps, ~9.4 million nested objects, 2.04 GB on disk. This is where the picture changes.

JSON Crack and JSON Hero are out of the running here. JSON Crack's documented ceiling is ~512 MB — past that, the graph layout and DOM cost don't survive. JSON Hero is designed for small payloads in a browser tab. That's not a knock — they're explicitly designed around smaller files.

Dadroit Standard handles up to 512 MB; for files this big you'd need its Advanced tier, which is marketed for files up to 1 TB. On Standard, 2 GB is past the supported ceiling. Where Advanced kicks in, the top of the tree is browsable quickly, and scrolling through deep nested arrays gets noticeably less snappy than on small files. Still respectable — most native viewers crater at this size.

JSONBolt first paint: still subsecond. The tape architecture means we only ever hold ~60 visible rows worth of materialized nodes regardless of total file size, and the mmap'd backing keeps RAM at roughly 1:1 with the file. The homepage shows a 1 GB cold-cache open in ~503 ms on an M2 Pro; you can read more about the engine in how to open large JSON files.

Test 3: searching for a value across 10 million records

We searched for a specific user-ID substring expected to match ~3,200 records inside the 2 GB file.

Web tools: not applicable. They never loaded the file.

Dadroit supports substring search and it works. On multi-gigabyte data we found the wait long enough to break interactive flow; for one-shot lookups that's fine, for exploratory drilling it adds friction. Your mileage will depend on disk and file shape.

JSONBolt runs substring search on its SIMD pipeline; the published homepage throughput is 20 M results/sec, with first-match latency on a 1 GB file shown at ~64 ms. On a 2 GB file the same query finishes well inside interactive-feel territory in our hands-on use. The GUI search bar has three toggles — case-sensitive, regex, and a key/value mode that lets you constrain the match to a specific property name. From the CLI, jb search adds --where predicates with comparison, regex, contains / startsWith / endsWith, length, and boolean composition, and accepts six different path syntaxes on input (jq, JSONPath, JSON Pointer, lodash, canonical, bracket-key). For piping into a larger workflow, see jq alternatives for large files.

Test 4: exporting a subset to CSV

Goal: pull every items[*] record from each top-level entry and emit a flat CSV.

JSON Crack: has a format converter (JSON ↔ CSV/YAML/XML) and code generators, so a one-shot conversion of a whole document is straightforward. Filtering a subtree first is not really its workflow.

JSON Hero: tree/column view is great for exploration; not built for bulk CSV export. Copy subtree, convert elsewhere.

Dadroit: JSON-to-CSV conversion is a built-in feature. Serviceable for flat data; complex nested paths or pre-filtering still want a step beforehand.

JSONBolt: the GUI ships a "Convert JSON to CSV and XML" step (it's in every tier, including free Personal). Drive the slicing from the CLI with jb extract '.items[*]' or jb search --where '...' --emit object, then open the result in the app and export. Details and edge cases in JSON to CSV for large files.

Where each tool actually shines

It would be cheap to declare a single winner. Here's what each one is genuinely the right pick for:

• JSON Crack — node-link diagrams. If you want to see the shape of a JSON document on a whiteboard or share an architectural diagram with someone non-technical, JSON Crack's graph view is the prettiest in the category. Open-source, self-hostable, and the team has been generous with the free tier.
• JSON Hero — small API responses. The column/schema view is the cleanest way we've seen to navigate a 500-line API response in a browser, and the public sharing flow is friction-free. If you're a backend engineer poking at someone else's API, this is a great daily driver.
• Dadroit — pragmatic native viewing. Mature product, three view modes (text, tree, table), built-in JSON-to-CSV, and the most generous free tier in the category: a no-cost non-commercial license across Windows, macOS and Linux. Commercial use is $98/yr (Standard, up to 512 MB) or the Advanced tier for files past that ceiling. If you want a no-fuss desktop viewer and you're a personal user, Dadroit is a totally reasonable pick.
• JSONBolt — gigabyte-to-terabyte files, SIMD search, CLI integration, NDJSON tail (live log tailing), and an --ai flag that turns a 4 GB blob into LLM-friendly slices (JSON and the LLM context window). If your workflow involves piping JSON between tools or feeding it to a model, this is what we built.

Honest concessions

Where JSONBolt is not the right answer:

• You want a graph diagram, code generation from a schema, or format conversion across JSON/YAML/XML/CSV. We render a fast virtual tree, not a node-link graph, and we don't generate TypeScript / Go / Rust types from a sample. Use JSON Crack.
• Personal-use forever, on every desktop OS, today. Dadroit has a free non-commercial license across Windows, macOS and Linux GA. Our free Personal tier is capped at 50 MB, and mac/Linux are still in private beta.
• You're on macOS or Linux and need GA-stable today. We're in private beta on those platforms; Dadroit is GA everywhere. Our Windows build is the GA one.
• You're inspecting a 50 KB API response. JSON Hero will be faster to reach for than installing anything.

Pick the right tool

A short decision rubric for the JSONBolt vs Dadroit / JSON Crack vs JSON Hero question:

Pricing for JSONBolt — free personal use under 50 MB, $80/yr Pro for unlimited and commercial, $80 lifetime while the limited tier lasts — is on the pricing page. No upsell screens, no usage telemetry, the free build is the full GUI with a file-size cap.