Skip to content
Changelog

Changelog

Tools shipped, features added, and changes made — newest first.

RSS feed
new tooltext & data

Word Counter counts words, characters, reading time, sentences, paragraphs, unique words and your longest word as you type or paste. It counts the way a person would, so emoji and Chinese, Japanese and Korean text are handled properly.

A keyword density panel ranks your ten most frequent words, skipping common English and Spanish filler words, with a toggle for short words. Hover or focus a keyword and every occurrence lights up in the text. Select part of the text and every stat switches to "n of total" for just that selection. Reading and speaking speeds are adjustable (default 225 and 150 words per minute), and they and your draft are remembered on this device.

A Readability card scores your writing with Flesch Reading Ease and Flesch-Kincaid Grade Level, with a band from "Very easy to read" to "Very difficult to read" and an approximate grade. Syllables are estimated from spelling, and a popover explains the formula and its limits. Both formulas suit English, so for very short text (under about two sentences or twenty words) or mostly non-Latin text the card explains instead of showing a number.

Copy summary gives you a plain-text block, which you can also download as .txt. Share puts the text in a link that restores it exactly. Counts update on every keystroke, while keyword density and readability catch up a moment later, so long drafts stay smooth.

new tooltext-data

SQL Formatter makes a pasted query readable. Pick a dialect (standard SQL, PostgreSQL, MySQL, MariaDB, SQLite, SQL Server T-SQL, BigQuery, Snowflake, Redshift or Oracle PL/SQL), a keyword case (upper, lower or as typed), an indent width, and how many blank lines to leave between statements.

It tidies joins, CTEs, window functions and whole migration files, and leaves string literals, -- and /* */ comments, placeholders (?, $1, :name, @name) and dialect syntax like Postgres :: casts or MySQL backticks exactly as you wrote them. Pasting formats automatically. Minify does the reverse and collapses a query to one line, safely around quoted strings and comments.

You can drag a .sql file onto the drop area, click to upload, or load the sample. Checking runs as you type, and the status says "No syntax errors" rather than "Valid SQL" because the check is lenient: it catches unclosed strings, unbalanced brackets and misplaced tokens, but it isn't a full grammar check. Errors show a line and column, and you can click to jump to them. Changing the dialect re-checks right away.

Your dialect, case, indent and statement-gap choices are remembered on this device, and Share puts the query and options in a link.

new tooltext & data

CSV Explorer turns a CSV or TSV file into a sortable, searchable table. Drop in a file or paste the text into the editor, or load the eight-column sample.

The delimiter (comma, semicolon, tab or pipe) and whether the first row is a header are detected for you, and you can override both from the toolbar. Each column gets a type (integer, number, date, boolean or string), and a Column info popover shows unique and empty counts, plus min, max and mean for numeric columns.

The table handles large files: a 200,000-row file scrolls as smoothly as a short one, and past that row count a notice in the status bar tells you the rest wasn't loaded. Wide files get a readable fixed column width, with the first column pinned while the rest scroll. Click a column header to sort ascending, descending, then back to unsorted, or use the Sort by controls in the toolbar.

The search box filters rows across all columns or just one. It is a plain, case-insensitive text match, so punctuation and brackets are matched literally, and every match is highlighted in place.

Whatever is visible, filtered and sorted, exports as CSV or JSON, or copies as JSON. A share link is offered for pasted text only. Open as JSON in JSON Formatter carries the visible rows across with the data already loaded.

new tooltext-data

YAML Formatter reformats a YAML document and converts between YAML and JSON. Paste or drop in your YAML and choose an indent of 2 or 4 spaces, a line-wrap width (0 turns wrapping off), and an optional alphabetical key sort.

Comments, anchors, aliases and merge keys (<<: *anchor) all survive formatting, and sorted keys keep their comments attached. Because a section-header comment can stop making sense once its key moves, the tool names the commented keys a sort would reorder and warns you before you rely on it. Multi-document files separated by --- are handled one document at a time, and several documents convert to a JSON array.

Validation runs as you type with line and column numbers. Duplicate keys show a warning rather than an error, and values like no, yes, on and off stay as strings instead of turning into booleans.

The swap control flips the tool to convert JSON into YAML, carrying the previous output across so a YAML to JSON to YAML round trip stays in one place. After converting, Work with this as JSON opens JSON Formatter and Query it with JSONPath opens JSONPath Explorer, with your converted output already loaded.

Your indent and line-width choices are remembered on this device, and Share puts the input and options in a link.

improvementjson

JSON Formatter can now open JSON sent from other tools. Links from YAML Formatter's Work with this as JSON open the formatter with your converted output already in the editor instead of an empty one. The data travels in the link itself, and it is cleared once loaded so a reload won't overwrite your edits.

new tooljson

JSONPath Explorer lets you paste a JSON document, type a JSONPath expression, and see every match as you type.

Each match shows its path (like $.store.book[0].title), which you can click to copy, and its value, collapsed when it's large. Filter expressions such as ?(@.price < 10) work, and a query can read your data but never run code.

The query field is a one-line editor with syntax highlighting and autocomplete built from your own document. Type a prefix like $.st to get the document's real paths, or press Ctrl/Cmd-Space to open suggestions. After .., @. or [ you get key names with their types, or wildcards, filters and slices.

Queries that matched something are kept in a History list, most recent first, up to 25, stored only on this device. Click one to run it again, remove entries one at a time, or clear them all. A Paths popover lets you browse the document's structure and click a path into the query, and a Cheatsheet covers recursive descent, wildcards, slices, unions and filters, with examples you can run against your document or the built-in bookstore sample.

Clicking a result's path jumps to that key or value in the JSON editor and selects it. If an edit has removed the match, the tool says so instead of jumping somewhere wrong. JSON errors and bad expressions are reported inline, each labeled so you know which side to fix. Results beyond 5,000 matches are capped with a notice.

An output panel gathers all matches as a JSON array you can copy or download, with a Pretty toggle. Share puts the query in the link, and the document too when it's small enough.

updatecss

The Flexbox Playground now has 17 one-click recipes, up from 5: Fill Space, Stretch Middle, HTML5, Alternating Grid, Masonry Column, Masonry Row, 3x3 Grid, 4x4 Grid, Align Center, Fill Right, Fill Center, Horizontal Bars, Vertical Bars, Row Wrap, Vertical Stack, Navbar and Sticky footer. The old Centered card, Equal columns and Tag cloud recipes are now Align Center, Fill Space and Row Wrap.

Each recipe is a card with a small diagram of its layout, in a grid you can move through with the arrow keys. The picker collapses behind a toggle, and it remembers whether you left it open. Recipes that need items of different sizes, such as the bar charts and masonry layouts, turn on "varied sizes" for you, and turn it off again when you switch to one that doesn't.

The preview now animates changes instead of snapping. Switching direction, loading a recipe with a different number of items, or editing grow or basis moves and resizes the items smoothly. Dragging still follows your pointer directly, and the animation is skipped if your system asks for reduced motion. Everything you build stays on your device.

new tooladvanced & local

Binary Inspector answers "what do these bytes mean?" Drop in a file or paste hex, choose an offset, and it decodes the bytes there as every common type in one table: u8 through i64, f32 and f64, unsigned and signed LEB128 varints, UTF-8 and UTF-16 strings, and two Unix timestamps (the 32-bit value as seconds and the 64-bit value as milliseconds). Each row copies in one click. A byte-order toggle re-decodes every multi-byte row as little- or big-endian. Near the end of a file, rows that don't fit say so.

A signature panel checks the start of the file against sixteen common formats, including PNG, JPEG, GIF, ZIP, PDF, ELF, WebAssembly, GZIP, BMP, Windows PE, Java class files, SQLite, RIFF, 7-Zip and Ogg. A hexdump of the 64 bytes around your offset highlights the exact bytes behind the current row.

It works together with Hex Editor. Select a byte in Hex Editor and a button opens it here with the surrounding bytes loaded, and a link here takes you back to edit. An arrow next to any of the eight integer rows opens the value in Base Converter with the matching bit width.

Only your offset and byte order are remembered between visits. The bytes you inspect are never saved or uploaded.

new toolimages

The Blur Placeholder Generator turns an image into a tiny blurred stand-in, the kind used for "photo fades in sharp" loading states.

It shrinks the image so its longest edge is between 4 and 32 pixels (a slider sets it), encodes it as WebP or JPEG, and the CSS blur scales it back up. The result is small enough to inline in your markup.

You get four outputs to copy: the base64 data URI, a CSS background-image plus filter: blur() rule, a next/image snippet with placeholder="blur" and the matching blurDataURL, and a plain <img> blur-up pattern for anywhere else. A side-by-side preview shows the placeholder scaled up next to the original, so you can judge whether 8px or 24px is enough.

Each placeholder also reports its average color, shown as a swatch with its hex value. The CSS and <img> snippets use it as a background-color fallback, so the first paint is a plausible color instead of a blank box. Every output shows its size in bytes, since a good placeholder should cost closer to a kilobyte than ten.

Your image is read in your browser and never uploaded.

new tooldeveloper

The CSP Builder is the second tool in the Developer & Web section. You can build a Content-Security-Policy from directive rows, or paste an existing policy to have it split apart and checked.

Build mode starts from a sensible defensive policy. Three presets cover common cases: Google's strict CSP (a per-response nonce plus 'strict-dynamic'), a locked-down static site, and a Next.js app with the allowances it needs. Anything that uses a nonce gets a real one generated when you click, and every row has its own Generate button. Loading a preset replaces the whole policy, and a toast lets you undo for a few seconds.

Each directive row has an editor suited to it. Source directives get one-click keyword chips for 'self', 'none', 'unsafe-inline', 'unsafe-eval' and 'strict-dynamic', common schemes like https:, data: and blob:, a field for hosts, and nonce and hash builders that write the correct 'nonce-…' or 'sha256-…' (or 384/512) syntax. report-to, report-uri, sandbox, trusted-types and require-trusted-types-for get token chips instead. An empty sandbox is kept, because it's the most restrictive form.

The output can be the enforcing Content-Security-Policy header, Content-Security-Policy-Report-Only, or a <meta http-equiv> tag. Report-only is the safe way to roll out a policy, and a report-only policy with no report-to or report-uri is flagged because it would collect nothing. The panel shows the size in bytes and warns past 4 KB, ahead of the 8 KB limit many servers apply to a header. With the meta tag, it warns when the policy has header-only directives like frame-ancestors or report-uri, because a meta tag ignores them.

Parse mode takes any policy (from a live site, a scanner or a message), splits it into directives and sources, and checks it. One click loads it into the builder so you can fix it. The checks cover 'unsafe-inline' and 'unsafe-eval', wildcard hosts, a missing object-src or base-uri, repeated directives (only the first applies), unrecognized names, and no script-src or default-src fallback. They also catch policies that mean the opposite of what they look like: 'strict-dynamic' with no nonce or hash, 'none' mixed with other sources, and http:// sources under upgrade-insecure-requests. 'unsafe-inline' next to a nonce is reported as a fallback for older browsers, not as a hole.

Warnings appear live in build mode, next to the row you changed. Share puts the whole policy and mode in a link, and your rows are remembered between visits.