Comparing two API responses to see what actually changed in a release, ignoring whitespace/key-order noise. Diffing config files before/after a migration to confirm only the intended fields moved. Generating an RFC 6902 JSON Patch document to ship to a system that supports it (PATCH endpoints, JSON-Merge-Patch fallbacks). Eyeballing two test fixtures to see what makes one fail when the other passes.

Array compare mode matters. "By index" reports an inserted element as remove+add for everything after it. "By value" treats arrays as sets, missing genuine reorderings. Pick the one that matches how your data is meant to be ordered. Number-vs-string isn't structural. {"id": 1} and {"id": "1"} show as a change because the types differ. Normalise types before diffing if that matters. RFC 6902 is a one-way patch, not a merge. Apply it with a real RFC 6902 implementation, not by string-replacement. Big trees get noisy. If the diff is hundreds of operations long, you're probably comparing two unrelated documents — re-check the inputs.

JSON Diff

Structural diff for two JSON documents — keys added, removed, changed, and value changes shown side by side.

Left (A)

Right (B)

Mode

Array compare

Result

Enter input above to see the result.

What is this for?

A plain-text diff on JSON tells you which lines changed; a structural diff tells you which data points changed. They're often very different — a re-formatted document with no semantic change is "every line different" to a text diff but "no changes" here. This tool walks both JSON trees and reports each path where they differ, using RFC 6901 JSON Pointer syntax (/users/0/name) so the output is unambiguous regardless of formatting.

When to use it

Comparing two API responses to see what actually changed in a release, ignoring whitespace/key-order noise.
Diffing config files before/after a migration to confirm only the intended fields moved.
Generating an RFC 6902 JSON Patch document to ship to a system that supports it (PATCH endpoints, JSON-Merge-Patch fallbacks).
Eyeballing two test fixtures to see what makes one fail when the other passes.

Common gotchas

Array compare mode matters. "By index" reports an inserted element as remove+add for everything after it. "By value" treats arrays as sets, missing genuine reorderings. Pick the one that matches how your data is meant to be ordered.
Number-vs-string isn't structural. {"id": 1} and {"id": "1"} show as a change because the types differ. Normalise types before diffing if that matters.
RFC 6902 is a one-way patch, not a merge. Apply it with a real RFC 6902 implementation, not by string-replacement.
Big trees get noisy. If the diff is hundreds of operations long, you're probably comparing two unrelated documents — re-check the inputs.