# JSON Error Locator Paste broken JSON and get the exact line, column and cause of every syntax error, with a caret-marked snippet and a suggested fix. Runs in your browser. ## Run it - **CLI:** `gizza tool json-error-locator '{ "name": "Ada", "tags": [1, 2,] }'` - **Web:** https://gizza.ai/tools/json-error-locator/ - **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/json-error-locator/tool.json ## Inputs - `json` — JSON to check _(field)_ - `output` — Result format _(field)_ - `context_lines` — Context lines around each error _(field)_ - `scan_all` — Report every issue (not just the first) _(field)_ ## Output - Diagnosis (text) ## Query parameters Open the tool pre-filled and auto-run via URL: - `json` — JSON to check - `output` — Result format - `context_lines` — Context lines around each error - `scan_all` — Report every issue (not just the first) Example: `https://gizza.ai/tools/json-error-locator/?json=%7B%0A%20%20%22name%22%3A%20%22Ada%22%2C%0A%20%20%22tags%22%3A%20%5B1%2C%202%2C%5D%0A%7D&output=report&context_lines=2&scan_all=true` --- ## Find where your JSON breaks, not just that it broke Parsers tell you *something* is wrong and stop at the first mistake. Paste the document above and this tool tells you **where** — a 1-based line, a character-counted column, and a 0-based offset for every syntax problem it finds — plus a plain-English cause, a concrete fix, and a caret pointing at the exact character. Everything runs in your browser; the text is never uploaded. ### What it names - **Trailing commas** — `[1, 2, ]`, `{"a": 1,}` - **Single-quoted strings and keys** — `{'a': 'b'}` - **Unquoted keys** — `{name: "Ada"}` — and unquoted bare-word values - **Missing commas** between members/elements, and a missing (or `=` instead of `:`) colon - **Mismatched or unclosed brackets** — `{"a": 1]`, `{"a": {"b": 1}` - **Unterminated strings**, pointing back at the opening quote that started them - **Invalid escapes** — `"C:\Users"`, and `\u` escapes with fewer than four hex digits - **Unescaped control characters** — a raw line break or tab inside a string - **Invalid numbers** — `.5`, `1.`, `+1`, `007`, `0x1F` - **JavaScript/Python literals** — `undefined`, `NaN`, `Infinity`, `True`, `None` - **Comments** — `// line` and `/* block */` - **Extra content** after the document's single top-level value ### Worked example Input: ``` { "name": "Ada", "tags": [1, 2,] } ``` Diagnosis (default settings): ``` Invalid JSON — 1 issue found. A parser stops at line 3, column 17. 1. Line 3, column 16 (offset 34) — trailing comma Cause: this comma is the last thing in the array; JSON does not allow a comma before ]. Fix: delete this comma (or add another value after it before the ]). 1 | { 2 | "name": "Ada", 3 | "tags": [1, 2,] | ^ 4 | } ``` Switch **Result format** to machine-readable JSON and the same run returns `{valid, issue_count, issues[], parser_stop, summary}` — one object per issue with `kind`, `label`, `line`, `column`, `offset`, `explain`, `fix` and the snippet — ready to pipe into a script or a CI check. ### Valid input still tells you something A document that parses returns a summary instead of an error: the top-level value's type, how many members or elements it holds, the nesting depth, and the line/byte count. ### Limits and edge cases - **Input is capped at 1 MiB** and nesting at **200 levels**; deeper input reports `nesting too deep`, which in practice almost always means an unclosed bracket further up. - **A position is where the problem was *detected*,** which can be just after the real mistake. For a missing comma or bracket, check the end of the previous line — the report repeats this note for exactly that reason. - **Columns count characters, not bytes**, matching what your editor's cursor shows. `"héllo"` is 6 bytes but 5 characters wide. - **An unterminated string ends the scan.** Everything after it was swallowed by the open quote, so any further finding would be an artefact rather than a real error. - **Syntax only.** This tool diagnoses; it does not rewrite your JSON and does not check it against a JSON Schema. - At most **50 issues** are listed — a badly mangled file cascades, and the first few are the ones worth fixing. ## FAQ
Why does the reported column point just after my mistake? Because that is where the document first stops making sense. A missing comma is only detectable at the *start of the next thing* — by the time `"b"` appears on line 3, the comma that should have ended line 2 is already missing. The same applies to unclosed brackets: the error surfaces at the end of the file, not where you forgot the `}`. The report flags the detection point and reminds you to look at the line above; for unclosed containers it points back at the opening bracket instead.
How is this different from a plain JSON validator? A validator runs a real parser, which aborts at the first error — a file with five mistakes takes five edit-and-retry rounds. This tool runs the parser *and* a tolerant scanner that keeps going after each problem, so all five are listed at once with individual positions, causes and fixes. Set **Report every issue** off to get the single-error, parser-style behaviour instead.
My JSON came from JavaScript or Python and nothing parses it. What now? That is the most common case here: single quotes, unquoted keys, `True`/`None`/`undefined`, `NaN`, and `//` comments are all valid in those languages and none of them are JSON. Each one is named individually with the exact replacement to make — for example `True` → `true`, or `name:` → `"name":`. Fix them top-down and re-run; the issue list shrinks as you go.
Is my data uploaded anywhere? No. The diagnosis is compiled to WebAssembly and runs entirely in your browser tab — no request carries your JSON off the page, so pasting a config file or an API response with credentials in it is safe. Closing the tab is all the cleanup there is.
What do the context-lines and offset controls do? **Context lines** sets how many source lines are printed above and below each flagged line, with a caret under the exact column — 2 by default, 0 to drop the snippets and get a compact list of positions, causes and fixes. The **offset** in each heading is a 0-based character offset from the start of the document, which is what `String.slice`, `seek`, and most editor "go to position" commands want.
Can it fix the JSON for me? Not here — this tool's job is the diagnosis, so you can see and understand every problem before changing anything. Repairing malformed JSON, and re-indenting JSON that is already valid, are separate tools; use the machine-readable output if you want to drive a fix from a script.
## Related tools - [JSON Repair](https://gizza.ai/tools/json-repair/): Fix broken JSON automatically — trailing commas, single quotes, unquoted keys, missing commas, comments, truncated LLM output. Runs in your browser. - [Absolute value, sign, or negation for a whole column](https://gizza.ai/tools/absolute-value-transformer/): Paste a column of numbers and apply absolute value, sign extraction (-1/0/1), sign flipping, or force-negative to every value at once, with rounding and an audit table. - [Adjacency Matrix Converter](https://gizza.ai/tools/adjacency-matrix-converter/): Convert a graph between edge list, adjacency matrix, and incidence matrix — directed or undirected, weighted or not. Free, private, runs in your browser. - [Amazon Order Analyzer](https://gizza.ai/tools/amazon-order-analyzer/): Paste an Amazon order-history CSV export to summarize total spend by month, top items, and category breakdowns. Browser-only, private, with Markdown or JSON output. - [ARFF Converter](https://gizza.ai/tools/arff-converter/): Convert Weka ARFF datasets to CSV and CSV tables back to ARFF locally — nominal attributes, numeric types, dates, sparse rows, missing values, and type rows.