Paste JSON (⌘V) or drop a file to inspect it
Nothing you paste or drop ever leaves your browser.
Explore, format, and edit JSON with interactive Tree and Split views. Query with JSONPath, compare side-by-side diffs, auto-repair syntax errors, minify, process JSONL streams, and convert JSON to TypeScript, Python, Go, Rust, Java, C#, Swift, CSV, YAML, XML & SQL. 100% local-first — your data never leaves your browser.
Paste JSON (⌘V) or drop a file to inspect it
Nothing you paste or drop ever leaves your browser.
Most JSON tools do one thing — format, or diff, or convert to one language — and send your data through their server to do it. Devure JSON does all of it, entirely in your browser, in one place.
Worker-based parsing + a virtualized tree means multi-MB documents stay responsive instead of locking the tab.
Parsing, diffing, converting, schema validation, JWT decoding — all local. No account, no upload step, no server round-trip.
Draft-07/2019-09/2020-12 auto-detected, human-readable errors, infer-a-schema-from-JSON, save/load named schemas, a built-in regex tester.
Mixed-shape arrays merge into one correct type instead of duplicating interfaces — verified against quicktype-style test cases, not just the happy path.
Yes, entirely free, and no. Every tool on this site — the viewer, the formatter, JSONPath, the schema validator, the converters — runs inside a Web Worker in your browser tab. There's no upload step, because there's no server that receives your document. See /privacy for the exact architecture and the full list of what the small amount of local analytics does and doesn't record.
Yes — parsing runs in a Web Worker and the tree view is virtualized, so only the rows actually on screen get rendered. The /large-files page is built and tested specifically for multi-megabyte documents that choke most browser-based JSON viewers.
Draft-07, 2019-09, and 2020-12, auto-detected from the schema's own $schema field (or pick one explicitly). Errors are humanized — 'Expected string, received number' at /users/0/email, not raw Ajv output — and you can infer a starting schema directly from a JSON sample.
TypeScript, Python (Pydantic/dataclass/TypedDict), Go, Rust, Java, C#, Swift, and JavaScript for code generation; CSV, YAML, XML, TOML, and SQL for data/config formats. Each has its own dedicated page with live options and a downloadable file.
No, by design — it decodes the header and payload and surfaces every standard claim (exp, iat, nbf, iss, sub, aud) plus expiry/not-before status, but signature verification needs the signing key, which this tool never asks for. That's the same decode-only scope jwt.io uses.
No account, ever — there's nothing to sign up for. The tool works over a normal page load; a dedicated offline/installable mode is on the roadmap but not shipped yet.
Full keyboard reference: press ? anywhere in the app, or open the privacy page to see exactly how the local-first architecture works.