JSONPath Evaluator

A JSONPath query tool that runs entirely in your browser. Paste JSON, type a path, get matches and their canonical paths side by side. Live evaluation, zero round-trips, no telemetry.

What JSONPath is

JSONPath is a query language for JSON. It was proposed by Stefan Goessner in 2007, became the de facto way to address pieces of a JSON document in tooling like kubectl -o jsonpath, CloudWatch dashboards, and most monitoring UIs, and was finally standardized as RFC 9535 in 2024. Think of it as XPath for JSON — a path expression that selects zero or more nodes from a tree.

A JSONPath expression always starts with $ (the root). From there, you chain segments:

• .name or ['name'] — select a member of an object.
• [0] — index into an array (negative indices count from the end, RFC 9535 style).
• [*] or .* — wildcard: every child of an array or object.
• ..name — recursive descent: find every descendant with the given key, at any depth.
• [start:end:step] — array slice, with all three bounds optional.
• [?(<expr>)] — filter: keep only children where the expression is truthy.

Supported filter operators

Inside [?(…)], you can build predicates using the current node (@):

• Comparison: ==, !=, <, <=, >, >=
• Logical: &&, ||, !, with parentheses for grouping
• Paths: @.foo, @.foo.bar, @['weird key']
• Literals: numbers (42, 3.14, -7), quoted strings ('fiction', "fiction"), true, false, null
• Existence: a bare path like [?(@.isbn)] is truthy whenever the key is present and non-falsy

Example: find every fiction book under $10.

$.store.book[?(@.category == 'fiction' && @.price < 10)]

What's not supported (yet)

• Script expressions ([(@.length - 1)]) — standardized as an optional extension in RFC 9535, deliberately omitted here. If you need to index from the end, use [-1].
• Function extensions — length(), count(), match(), search(), value() are not parsed. A future release may add the core five.
• Unions — $.a[0,2,4] (index union) and $.a[(name == 'x'),(name == 'y')] are not supported. Evaluate multiple queries separately.
• Regex filters — there is no ~= or match() operator. Post-process with the regex tester if you need pattern matching.

How it compares

• vs JMESPath — JMESPath (used by AWS CLI --query) is a different language, not a superset. It has cleaner array projection syntax and a rich function library, but it is not what kubectl or CloudWatch speak. Pick your dialect based on the tool you're targeting.
• vs jq — jq is a full functional language that happens to use JSON-like syntax. It can do everything this tool does and much more, but it is a separate runtime you install. If you just need to extract a few fields while building a query to paste into a kubectl flag, this is faster.
• vs JSON.parse(x).a.b[0] — hand-written property access is faster for a single lookup but breaks on missing intermediate keys and doesn't handle wildcards or recursion. JSONPath is the expressive escape hatch.

Keyboard shortcuts

• ⌘/Ctrl + Enter — re-run the query and persist the state into the URL hash
• ⌘/Ctrl + K — open the global command palette (jump to another tool)

Privacy promise

Nothing you paste ever leaves your device. The page loads no analytics scripts, no third-party fonts, no trackers. Share links put state in the URL fragment, which by specification is never transmitted to a server. Even the "Copy matches" / "Copy paths" buttons run entirely in-page via the Clipboard API.

Performance notes

The evaluator caps recursion at depth 100 to protect against pathological documents. Wildcards and recursive descent traverse every descendant — on very large documents (hundreds of MB), this is O(n) in the document size. If you hit a slow query, pre-filter the document with a narrower path before running a .. descent.