Every tool on one bench. Learn them by using them.
Hushbench is a private data workbench: formats, spreadsheets, SQL, diagrams, logs and PII checks that run entirely on your machine. This guide works the same way. Edit any example below and watch it react, then open the real tool.
- Alltools in one app
- 0 bytesof your data uploaded
- 1 GBlargest file it opens
Getting started
One workbench instead of a dozen browser tabs
Most of us paste payloads, tokens and logs into whichever online formatter comes up first. Hushbench replaces those sites with one app that never sends your data anywhere. Pick a tool to jump straight to its section.
Getting started
Use it in the browser or on your desktop
Open Hushbench on the web or install the desktop app. Your work stays on your device either way, and your license works in both.
In the browser
Open hushbench.msweelam.dev. Nothing to install. Works in current Chrome, Edge, Firefox and Safari.
The page needs a connection to load; after that, everything you paste or open is processed in your browser.
Desktop app
Native builds for macOS (DMG, ZIP), Windows (installer, portable EXE) and Linux (AppImage, DEB).
Works fully offline once installed, which suits restricted networks. Adds a native menu, Help → Manage License and Help → Documentation.
Getting started
What stays on your machine (everything you work on)
Parsing, conversion, queries, diagrams and scans all run in your browser or desktop app. There is no Hushbench server that sees your data, so there is nothing to leak, log or subpoena.
Your content
- Never uploaded, in any tool.
- Session state is kept in your browser's local storage and cleared after 7 days.
- Databases open in memory; the original file is never changed.
The hosted website
- Counts visits with one anonymous cookie, so we know how many people use it.
- No third-party analytics or ad scripts.
- The desktop app makes no such request.
When the network is involved
- Web Preview runs the code you write; if it loads a CDN or calls
fetch, that request goes out. - The PII AI model (about 137 MB) is fetched once from the same site, then runs locally.
- Pro subscriptions refresh the license file when you happen to be online.
Getting started
Keyboard first, mouse optional
Every tool shares the same workspace. Panels resize by dragging the divider, and there are 15 themes (9 dark, 6 light) in the header. This page has its own quick jump too: press ⌘ K or /.
Getting started
Built for the 800 MB log file, too
Hushbench picks a strategy by file size. Small files render instantly; medium files are parsed in a background worker so the page stays responsive; large files ask before loading, then stream in 1 MB chunks and render only what's on screen (about 10,000 lines or 5,000 tree nodes at a time).
| Format | Instant | Background worker | Maximum |
|---|---|---|---|
| JSON | up to 5 MB | up to 50 MB | 1 GB |
| XML | up to 10 MB | up to 100 MB | 1 GB |
| Protobuf | up to 10 MB | up to 100 MB | 1 GB |
| YAML | up to 2 MB | up to 20 MB | 500 MB |
| Log Tracer | up to 10 MB | up to 50 MB | 500 MB |
| Markdown | up to 5 MB | up to 50 MB | 200 MB |
Tools
Structured Data
Read, fix, reshape and explain the documents you deal with every day.
JSON
Paste a payload and get a readable, validated document straight away. Switch between code, tree and table views, query it with JSONPath, diff two versions, or check it against a schema. Files up to 1 GB, all on your device.
- Paste to format. Pasting into the Code view formats the document immediately.
- Table view. Arrays of objects become a grid you can sort, filter and export to CSV or TSV.
- JSONPath queries with one-click examples like
$..authorand$.store.book[?(@.price < 10)]. - Diff / Compare shows Original, Modified and the Diff Result side by side.
- Schema Validation with quick templates for objects, arrays and OpenAPI.
- Generate models as a Java POJO, C# POCO, Python dataclass or Go struct.
How to use it
- Open Structured Data ▸ JSON.
- Paste, click Load File, or drop a file onto the editor.
- Pick a view from View: Code, Table or Tree. For checks, use Analyze: Validate, Diff / Compare, JSONPath Query or Schema Validation.
- Need code? Choose a language from Generate.
- Copy or Download the result.
Good to know
- Views
- Code · Table · Tree · Minify
- Analyze
- Validate · Diff · JSONPath · Schema
- Indent
- 2 spaces · 4 spaces · Tab
- File size
- Up to 1 GB
XML
Pretty-print or minify XML, check that it is well-formed, and explore it as a tree. Namespaces, CDATA, comments and right-to-left text all come through intact.
- Format / Prettify with 2 spaces, 4 spaces or tabs, or Minify for transport.
- Validate reports the first well-formedness error and where it is.
- Tree View to fold and browse large documents.
- Generate model classes from a sample document, in the same four languages as JSON.
How to use it
- Open Structured Data ▸ XML.
- Paste XML or click Load File.
- Choose Format / Prettify, Minify, Validate or Tree View.
- Copy or download the output.
Good to know
- Validation
- Well-formedness. XSD and DTD schemas are not checked.
- File size
- Up to 1 GB
- Convert
- To JSON or YAML with Universal Conversion
YAML
One misplaced space can break a deployment. The YAML tool tells you which line is wrong and why, formats the file consistently, and converts it to JSON when another tool needs it.
- Errors that explain themselves: line, cause and a suggested fix.
- YAML → JSON conversion in one click.
- Tree View for deep Kubernetes and CI configs.
- Consistent formatting with 2- or 4-space indentation.
How to use it
- Open Structured Data ▸ YAML.
- Paste or load a file.
- Pick Format / Prettify, Validate, YAML → JSON or Tree View.
- Fix any reported line, then copy or download.
Good to know
- Documents
- One per input. Multi-document files separated by
---are not split. - File size
- Up to 500 MB
Universal Conversion
JSON, YAML and XML in one workspace. Paste anything, leave From on Auto, choose a target, and the result follows as you type.
- Auto-detects the source format.
- Live: converts on every keystroke. ⇄ swaps source and result.
- Predictable XML: attributes map to
attr_<name>keys, text beside attributes maps tovalue. - Lists become repeated elements (or
<item>), wrapped in<root>when needed.
How to use it
- Open Structured Data ▸ Universal Conversion.
- Paste your data, or pick an example (JSON API response, YAML config, XML feed).
- Leave From on Auto and choose To: JSON, YAML or XML.
- Copy or download the result.
Good to know
- From
- Auto · JSON · YAML · XML
- To
- JSON · YAML · XML
- Indent
- 2 or 4 spaces
Smart Diagram
Turn a JSON, YAML or XML document into an interactive graph of cards. Handy for explaining a payload in a review, or for finding where a field lives in a 3,000-line response.
- Two layouts: Left to Right or Top to Bottom, laid out automatically.
- Double-click a card to collapse or expand its branch.
- Click a card and the matching code is highlighted in the editor.
- Personal data stands out: cards holding emails or phone numbers get an amber border.
- Search nodes, plus a minimap once the graph passes 20 cards.
- Long arrays are trimmed to 20 rows with a "+N more" line.
How to use it
- Open Structured Data ▸ Smart Diagram.
- Paste data or Load Example… (User Registration, Service manifest, Order feed).
- Pick a Layout; the format is detected for you.
- Use Fit to View, search, and double-click to fold branches.
Good to know
- Formats
- JSON · YAML · XML
- Zoomed out
- Below 50% zoom, rows turn into skeleton bars to keep it fast
Tools
Encode & Decode
Tokens, encodings and binary messages, handled where nobody else can see them.
Transform Pipeline
Chain encoding, hashing and encryption steps and see every intermediate result. Unwrap a cookie, prepare a token, or encrypt a secret before you hand it over, all offline.
- Encode & decode: Base64, Base64url, Hex and URL.
- Hash: SHA-256, SHA-384 and SHA-512.
- AES-256-GCM encrypt and decrypt with a passphrase (PBKDF2-SHA-256, 100,000 rounds).
- PBKDF2 derive key with your own salt, iterations and key length.
- JWT decode, offered automatically when the input looks like a token.
- Re-runs as you type, and errors name the step that failed.
How to use it
- Open Encode & Decode ▸ Transform Pipeline.
- Paste your text into Input.
- Add steps under Pipeline; AES steps ask for a Passphrase.
- Read the Output and click Copy. Clear chain starts over.
Good to know
- Samples
- Decode a JWT · Base64 round-trip · SHA-256 hash
- Crypto
- The browser's built-in Web Crypto API
- JWT
- Decodes only; signatures are not verified
Base64
Encode or decode text and files: standard, URL-safe, or a complete data URL ready to paste into HTML or CSS.
- UTF-8 correct: emoji, Arabic and accented text survive the round trip.
- Files too: drop any file to encode it as binary, with progress for files over 5 MB.
- Image preview when the file is an image.
- Decode to a file: Download writes the original bytes, with the extension taken from the data URL.
How to use it
- Open Encode & Decode ▸ Base64.
- Pick Encode or Decode, then a format: Standard, URL-Safe or Data URL.
- Type, paste, or drop a file.
- Copy the result or download it as a file.
Good to know
- File limit
- 50 MB per file
JWT
A production token is a credential. Decode it here instead of pasting it into a website: read the header and claims, see when it expires, and catch common mistakes.
- Decode header, payload and signature, with time claims shown as dates.
- Validate flags a missing or unknown
alg, unsecurednonetokens, an oddtypand expired tokens. - Auto-repair: a token whose payload and signature were glued together is split again, with a warning.
- Knows HS, RS, ES and PS algorithms at 256, 384 and 512 bits.
How to use it
- Open Encode & Decode ▸ JWT.
- Paste the token, with or without the
Bearerprefix. - Use Decode to read it, Validate to check it, or Analyze for detail.
Good to know
- Modes
- Decode · Validate · Analyze
- Not covered
- Signature verification, and encrypted JWE tokens (five parts)
Protobuf
Decode protobuf binaries against your .proto, or go the other way and encode JSON to binary. Useful for gRPC payloads, queue messages and cached blobs.
- Decode Binary from Base64, hex or a file into readable JSON.
- Encode to Binary: Generate JSON builds a starter message from the schema.
- Schema View analyzes your
.protoas you edit it. - Nested messages, repeated fields, enums and oneofs.
How to use it
- Open Encode & Decode ▸ Protobuf.
- Paste your schema or click Load Schema.
- Paste the binary as Base64 or hex (or load a file) and choose the Message Type.
- Read the JSON output, or switch to Encode to Binary.
Good to know
- Modes
- Decode Binary · Encode to Binary · Schema View
- File size
- Up to 1 GB
Tools
Docs & Sheets
Write, draw, calculate and prototype without leaving the bench.
Mermaid
Diagrams as code that live next to your docs and diff cleanly in git. Write the text, get the picture, then zoom, present or export it.
- 10 diagram types: flowchart, sequence, class, state, ER, user journey, Gantt, pie, git graph and C4.
- Zoom from 10% to 500% toward your pointer, plus Fullscreen and Preview Only.
- Export to SVG, or PNG at twice the resolution.
- Open in Atlas Whiteboard turns the diagram into shapes you can rearrange.
How to use it
- Open Docs & Sheets ▸ Mermaid.
- Write a diagram, or Load Example… for any of the 10 types.
- Zoom with the toolbar or your scroll wheel; Fullscreen for presenting (Esc exits).
- Export as SVG or PNG.
Good to know
- This demo
- Draws flowcharts only; the workbench renders full Mermaid
Atlas Whiteboard
An infinite canvas for architecture sketches and design sessions, built on Excalidraw, with engineering figures and Mermaid import. Your boards stay on your machine.
- Single-key tools: B rectangle, O circle, D diamond, K connector, T text, F frame, L laser.
- Figures panel: Database, Queue, Server and Actor, plus basic shapes.
- Import Mermaid and get editable shapes.
- Export PNG at 1×, 2× or 3×, SVG, or to the clipboard; save as
.excalidraw. - Autosaves locally as you work. Press ? for every shortcut.
How to use it
- Open Docs & Sheets ▸ Atlas Whiteboard and name the board.
- Pick a tool or drag a figure onto the canvas.
- Connect shapes with the connector (K).
- Export an image, or Save the board to share it.
Good to know
- Keys
- ⌘ on macOS, Ctrl on Windows and Linux
- Files
- Open ⌘O · Save ⌘S · Export ⌘⇧E
- Editing
- Undo ⌘Z · Duplicate ⌘D · Group ⌘G
- View
- Reset zoom ⌘0 · Full screen ⌘⇧P
Markdown
A live GitHub-flavored preview with diagrams, math and images, and export to HTML or a styled PDF that is ready to send.
- Mermaid fences render inline, up to 100 diagrams per document.
- Math with KaTeX:
$…$,$$…$$,\(…\)and\[…\]. - Images: paste, drop or insert them (up to 5 MB each); they are embedded in the document.
- PDF export (Pro): 6 templates, page sizes from A6 to A0 plus Letter and Legal, cover page, running header, page numbers and watermarks.
- Copy HTML or download a standalone
.htmlfile.
How to use it
- Open Docs & Sheets ▸ Markdown.
- Write, or Load File. Choose Side-by-Side, Preview Only or Edit Only.
- Add images with Insert Image or paste them.
- Export with Copy HTML, Download .html or Download PDF.
Good to know
- PDF templates
- Classic Report · Modern Minimal · Technical Whitepaper · Executive Brief · Academic Paper · Midnight
- File size
- Up to 200 MB
Spreadsheet
Open CSV, TSV, XLSX, XLS, ODS and Numbers files and work with them the way you would in a desktop spreadsheet (formulas, filters, charts) without uploading the file anywhere.
- Table Preview for reading; Edit Grid for real spreadsheet work.
- Formulas including the
*IFSfamily andXLOOKUP. - Data panels: Group by, Chart (bar, line, pie, doughnut), Formula column, Conditional formatting.
- Clean-up tools: Remove duplicates and Text to columns.
- Several sheets per workbook; export to XLSX, XLSB, XLS, ODS, CSV, TSV, JSON or XML.
- Excel-style keys: fill handle, freeze panes, F2, Ctrl+D, Ctrl+; and 20 more.
How to use it
- Open Docs & Sheets ▸ Spreadsheet.
- Use File ▸ Open file…, Paste data, or start a New blank sheet.
- Read in Table Preview, or switch to Edit Grid to change cells and add formulas.
- Use the Data menu for grouping, charts and clean-up.
- Export from the File menu. JSON and XML exports can include only the filtered rows.
Good to know
- Delimiters
- Auto-detect · comma · tab · semicolon · pipe
- Zoom
- 50% to 200%
- Not yet
- Pivot tables. References to other sheets show
#REF!.
Web Preview
An HTML, CSS and JavaScript playground with a live preview and a console. Test a snippet, reproduce a bug, or check an email template before you send it.
- Three views: Side by Side, Code and Preview, with optional CSS and JS panes.
- Live: re-renders half a second after you stop typing.
- Console with All, Logs, Warnings and Errors filters, including failed resource loads.
- Sandboxed: your code cannot reach Hushbench or your other data.
How to use it
- Open Docs & Sheets ▸ Web Preview.
- Write HTML, or Load Example… (Dessert Break, CDN + fetch demo).
- Add + CSS or + JS panes as needed and watch the console.
- Format, Copy HTML or Download .html.
Good to know
- Network
- Code you write can load from the internet; the app shows a "Works now, but not offline" hint when it does
- Scope
- Single page; no npm packages or bundling
Tools
Ops & Infrastructure
Incident logs and deployment files stay sensitive. Work on them locally.
Log Tracer
Drop raw logs from several services and get traces back: a trace list, a log-level dashboard and a waterfall timeline for each request. No tracing backend required.
- Formats: Sleuth / OTel, JSON, Log4j, or your own regex with named groups.
- Auto-detects W3C Trace Context, logfmt, Elastic ECS, B3 and Jaeger headers.
- Traces panel: search and sort by duration, log count or first seen.
- Log levels: a donut chart; click a level to filter the source.
- Timeline: zoom in and open any span's logs in a drawer.
- Parses in a background worker; big sources scroll smoothly.
How to use it
- Open Log Tracer.
- Drop a
.log,.txt,.ndjsonor.jsonfile, or paste logs. - Pick a format template, or leave the profile on Auto-Detect.
- Click Parse Logs (⌘/Ctrl+Enter), then pick a trace.
Good to know
- Custom
- Regex groups:
traceId(required),timestamp,service,spanId,message - Large files
- Asks before loading files over 50 MB
Docker Compose
Validate a compose file, see how its services depend on each other, and turn it into a Helm chart when you move to Kubernetes.
- Four-layer validation: YAML syntax, compose structure, service settings, and references between services.
- Visualize draws the service dependency map.
- Helm Convert generates
Chart.yaml,values.yaml, helpers, notes, and a Deployment, Service and volume claim per item. - Honest conversion notes for
depends_on,network_mode, build contexts and bind mounts.
How to use it
- Open Infra-as-Code ▸ Docker Compose.
- Paste your file, or Load Example… (Simple Web + DB, Full Stack App, Microservices).
- Run Validate and fix what it reports.
- Switch to Visualize or Helm Convert; copy files one by one or download them all.
Good to know
- Services
- A Kubernetes Service is generated only for compose services that publish ports
- Bind mounts
- Become
emptyDirvolumes - Download
- All files as one
helm-chart-files.txt
Helm Templates
Render templates against YAML or JSON values and see the final manifest before it reaches a cluster. It works offline and needs no CLI.
- Handlebars syntax:
{{app.name}},{{#if}},{{#each}}(orrange),{{#with}},{{#unless}}. - Helpers:
default,upper,lower,quoteandtrim. - Scaffold builds a values file from the variables in your template.
- Validation Errors drawer, render time and a count of detected variables.
How to use it
- Open Infra-as-Code ▸ Helm Templates.
- Write a template or Load Example… (K8s Deployment, ConfigMap, Service, Ingress).
- Choose YAML Values or JSON Values, or click Scaffold.
- Copy or download the Rendered Output.
Good to know
- Shortcuts
- Open template ⌘O · Download output ⌘S · Scaffold ⌘⇧S
- Syntax
- Paths look like
{{app.name}}, not Go-style{{ .Values.app.name }}
Tools
SQL Studio
Design schemas and query real databases with engines compiled to WebAssembly. No server, no upload.
Schema Designer
Paste DDL and get an ER diagram you can rearrange. Change the SQL and the diagram follows; translate the schema between SQL dialects in the browser.
- SQL Conversion between PostgreSQL, MySQL, SQLite, T-SQL and generic SQL.
- Schema Diagram (Sync): the SQL editor and the diagram stay in step both ways.
- Diagram Edit (Layout): drag tables; they snap and align, and editing SQL never moves them.
- Hover a table to highlight its relationships, drawn in crow's-foot notation.
- Export the diagram as PNG, or the converted schema as
converted.sql.
How to use it
- Open SQL Studio ▸ Schema Designer.
- Paste
CREATE TABLEstatements or load an example (E-Commerce, Blog, Users' Logistics). - For a dialect change, use SQL Conversion and set From and To.
- For the picture, use Schema Diagram; switch to Diagram Edit to arrange it, then Auto layout or Fit.
- Export PNG or download the SQL.
Good to know
- Dialects
- PostgreSQL · MySQL · SQLite · T-SQL · Generic
- Engine
- polyglot-sql, running as WebAssembly in the page
SQL Explorer
Open a SQLite file or a PostgreSQL dump and query it in the browser, using the real engines compiled to WebAssembly. No database client to install, and the file never leaves your machine.
- Two engines: official SQLite (sqlite-wasm) and PostgreSQL (PGlite), each in a background worker.
- Schema tree of tables, views and triggers (and enums for Postgres), with a filter.
- SQL editor with highlighting and autocomplete for your own tables and columns.
- Results you can sort, page through and export to CSV or JSON.
- Schema DDL tab with the full
CREATEscript, ready to copy. - Samples to practice on, such as window functions on Orders and JSONB on Users.
How to use it
- Open SQL Studio ▸ SQL Explorer and pick a DB Engine.
- Open database (
.db,.sqlite,.sqlite3,.sql), start a New DB, or Load sample…. - Write a query and press ⌘/Ctrl+Enter.
- Sort and page the results, then export them.
Good to know
- Results
- Up to 10,000 rows, 100 per page
- Changes
- Run in memory; your original file is never modified
Tools
Security
Check what you are about to share before you share it.
PII & Secrets
Before a log goes into a ticket or a config goes into a chat, scan it for personal data and leaked credentials. Two layers of detection, both running on your device.
- Pattern layer: emails, phone numbers (including Saudi mobiles), SSN-like and national IDs, IBANs, and card numbers confirmed with the Luhn check.
- Secrets: AWS, GitHub, GitLab, Slack, Stripe and Azure keys, JWTs, PEM private keys and generic API keys, plus the secretlint rule set.
- Entropy check catches random-looking tokens that no rule knows.
- Smart Detection: an on-device AI model finds names, places, organizations, addresses, passports and account numbers. It runs only when you click it.
- Arabic-Indic digits are normalized, so ٠٥٥١٢٣٤٥٦٧ is caught too.
- Filter by type and severity, then View Report or Download MD.
How to use it
- Open PII & Secrets.
- Paste text or load a file. Pattern matches appear immediately.
- Click Run Smart Detection and confirm, to add the AI layer.
- Filter the findings, then View Report or Download MD.
Good to know
- Smart Detection
- Runs on the website and in the desktop app; uses WebGPU when available
- Model
- About 137 MB, downloaded once
- AI limit
- Files up to 2 MB; larger files use the pattern layer only
Plans
Free for everyday formats, Pro for the heavy tools
One build of Hushbench, two tiers. The free workbench has no time limit and no sign-in. Pro unlocks the advanced tools and Markdown PDF export with a signed license file that is checked on your device.
Free
All the everyday format tools, on the web and in the desktop app.
Pro
Adds the advanced tools and Markdown PDF export. Team and Enterprise plans cover organizations and air-gapped networks.
Help
Questions people ask
Is any of my data uploaded?
No. Every tool works on your device. The hosted website counts visits anonymously, but it never sees what you paste, open or produce.
Does it work without internet?
The desktop app works fully offline once it is installed. The website needs a connection to load, and from then on every tool processes your data inside the browser. The only exceptions are network-related by nature: code you run in Web Preview that loads remote resources, and downloading the PII AI model the first time.
Where are my settings and recent work kept?
In your browser's local storage, on your device. Each tool restores its last session, and saved sessions expire after 7 days. Whiteboards autosave separately.
Which browsers are supported?
Current versions of Chrome, Edge, Firefox and Safari. Smart Detection in PII & Secrets is fastest in browsers with WebGPU.
I found a bug or want a feature. Where do I go?
Open an issue on the issue tracker, or email hushbench.support@gmail.com.