Run JSONPath queries in your browser, see the matched values and copy any node's path in five dialects.
This tool does two things that usually live apart. First, you write a JSONPath query and it is actually evaluated against your document: the output is the list of matched values, each with the concrete path that reaches it, with a copy button per value and one for the whole set. Second, you walk the JSON tree, opening only the branches you care about, and clicking a node turns its path into the query, which is the reverse flow and the most common one day to day: you already found the value on screen and want its address. The selected path is shown in five dialects: JSONPath, dot notation, JavaScript accessor, Python accessor and jq filter. Errors are explicit: broken JSON reports line and column along with the offending line, and a malformed query reports the exact position of the problem and what was expected there, instead of returning an empty list and leaving you guessing. Everything happens in the browser, including files you drop into the input box.
JSONPath never had a single specification at the outset, and each library implemented a slightly different slice of it. That is why it is worth stating exactly what is evaluated here instead of promising "full JSONPath". The engine covers the root, child access by dot and by bracket, indexes (including negative ones, counted from the end), slices with a step, the wildcard, recursive descent, unions, and filters with comparison, existence and logical operators.
Inside a filter, @ is the item under test and $ is the document root, which lets you compare a field with a value that lives somewhere else, as in $..book[?(@.price < $.expensive)]. @.field.length also works, for the size of arrays and strings. When a query uses something outside that list, the tool raises an error with a position instead of silently returning zero results.
Much of the real work is not writing the query, it is finding the address of a field you are already looking at in a payload. The tree solves that: it opens on demand, so large documents do not freeze the tab, and every node you click fills the query with its concrete path. Searching by key or value is the same shortcut for when you know the field name but not where it is buried.
With a node selected, the same path appears in five writings. JSONPath is what testing and transformation tooling expects; dot notation is what utilities such as _.get take; the JavaScript and Python accessors paste straight into code; and the jq expression is for the command line. Keys with a space, a hyphen or a symbol come out escaped the right way for each dialect, which is exactly where a path written from memory tends to break.
Script expressions, such as the classic [(@.length-1)], are not executed: running arbitrary user code to resolve a path is an unnecessary risk, and the use case is almost always "the last item", which here is written [-1]. Regular expression comparison, aggregation functions and slices with a negative step are also out. In all of those cases you get a message saying what is unsupported and where, which is more useful than an empty list.
Keep in mind that engines genuinely diverge on details: some apply filters only to arrays, others to objects too (here, to both); some order recursive descent results differently; some accept unquoted keys inside brackets (so does this one). Before pinning a query in production, confirm the behavior in the library your project uses. To keep the page responsive, result and node lists have a cap, flagged when it is reached.
Paste the code into your HTML and the tool shows up on your page, without J-Kit's navigation and ads. It still runs in the browser of whoever visits your site.
<iframe
src="https://jkit.tools/embed/en-US/json-path-finder"
width="100%"
height="600"
style="border:0"
loading="lazy"
title="JSON Path Finder"
></iframe>These references help contextualize formulas, standards, APIs and limitations used on this page. They do not replace professional validation when a result has legal, financial, medical or operational impact.
Yes. You write the query and it is evaluated against the JSON you pasted, returning the list of matched values and the concrete path of each one. The engine covers the root, child access by dot and bracket, indexes (including negative ones), slices with a step, the wildcard, recursive descent, unions, and filters with comparison, existence and logical operators. It all runs in the browser, with the document never sent to a server.
4 matched values
"Nigel Rees""Evelyn Waugh""Herman Melville""J. R. R. Tolkien"$objobj.JSON stays in your browser. Files opened here are never uploaded, and neither queries nor values reach analytics.