A disk-space visualizer for Windows that scans a 2-million-file system drive in about 3.5 seconds.
Sprawl scans a drive or folder and draws it as a treemap. Each rectangle's area matches the bytes it uses, so the biggest space hogs are the biggest blocks. A sorted list beside the map gives exact sizes. As administrator, Sprawl reads the NTFS master file table directly instead of opening every folder, which is where the speed comes from.
Sprawl is built with Tauri 2, Rust and React.
The GUI supports English, French, Spanish and German. Choose a language in
Settings, or use the system default. Translations live in src/locales/ and
are shared by the frontend and backend. To add a language, add its JSON catalog
and register it in src/i18n.ts and src-tauri/src/i18n.rs. English is the
source of truth and fallback. The CLI stays in English.
- Fast scans. As administrator, Sprawl reads whole NTFS drives straight from the master file table. Without admin, it walks folders and drives in parallel. Drive rescans try to replay the NTFS change journal and only fall back to a full read when they can't.
- Treemap. The map shows the current folder three levels deep, coloured by file type or age, and stays sharp on high-DPI screens. Hover a block for its size, size on disk and item counts. Click a folder to zoom in. Right-click for actions.
- Size on disk. Next to each logical size, Sprawl shows the space the drive actually allocates, so compressed, sparse and cloud-placeholder files show their real usage. MFT scans count a hard-linked file's space once.
- Sorted contents, largest files and type breakdown. Each folder lists its children by size and percentage, the 50 largest files under it, and how much space goes to Video, Images, Audio, Documents, Archives, Code, Apps & system and Other. Hover a category to highlight its files in the map.
- Search. Find items under the current folder by name, wildcard, size,
type, kind or modified date/age, such as
*.iso size>1GBorage>1y. The side panel lists the largest matches, and the map dims everything else. - Export. Save every file and folder under the current folder, with full paths, sizes, size on disk, last-modified times, item counts and type, as CSV or JSON.
- All files and extensions. Browse every file with filtering, sorting by size, name or modified date, and Show more paging. The Extensions tab shows totals for each extension; click one to search for its files. CLI scan/show reports also include extension totals.
- Changes since last scan. Sprawl saves each completed scan. When you scan the same path again, it shows how much each folder grew or shrank, which items are new, what disappeared, and the 50 files that grew the most.
- Saved scans. The home screen reopens the last scan of any drive or recent folder without rescanning. Saved results show their age and warn you once they look stale.
- Missing space. At a drive root, Sprawl reports how much used space the scan couldn't account for, comparing it with the size on disk it found. System files, restore points and unreadable folders usually explain the gap.
- Safe deletion. You can send an item to the Recycle Bin, or delete it permanently after a second confirmation. Sprawl refuses to delete drive roots, the scan root, the Windows and Program Files folders, ProgramData and your user profile folder. It never opens or runs files. It can only reveal them in Explorer.
- Explorer integration. Add Scan with Sprawl to the right-click menu for folders, folder backgrounds and drives. This doesn't need admin.
- Keyboard shortcuts. Backspace or Alt+Left goes back, Alt+Up goes to the parent, Ctrl+F focuses the search, Delete opens the delete dialog, Shift+F10 opens a context menu, and Escape clears the search and highlights.
- Light and dark themes. Sprawl follows the Windows setting.
I ran each tool three times on the same machine against the same drive, with default settings. Lower is better.
Full drive scan of C:\ as administrator. The drive holds about 2.08 million
files and 549,000 folders, 980 GB in total.
| Tool | Median | Runs |
|---|---|---|
| Sprawl 0.1.0, MFT read | 3.5 s | 3.4, 3.5, 4.4 s |
| WizTree 4.14 | 10.5 s | 9.5, 10.5, 11.7 s |
| WinDirStat 2.8.0 | 13.3 s | 13.2, 13.3, 16.6 s |
| Sprawl 0.1.0, directory walk without MFT | 15.2 s | 6 runs, 11.0 to 28.6 s |
Folder scan of C:\Windows, about 199,000 files and 141,000 folders.
| Tool | Median | Runs |
|---|---|---|
| Sprawl 0.1.0 | 2.7 s | 2.6, 2.7, 2.7 s |
| WinDirStat 2.8.0 | 8.4 s | 8.3, 8.4, 8.5 s |
| WizTree 4.14 | 31.1 s | 31.1, 31.1, 34.7 s |
WizTree is fast on a whole drive but took three times longer on a single
folder than on all of C:\. If you mostly scan folders, that matters.
The test machine has an AMD Ryzen 9 7900X3D, 64 GB of RAM and a 1 TB Sabrent NVMe SSD, running Windows 11 Home build 26200. Every tool ran elevated, one at a time, after a warm-up scan.
The three tools weren't timed the same way, so treat the tables as a rough comparison.
- Sprawl. The time is what Sprawl's own timing tests print for the scan itself. App startup and drawing the first treemap aren't included.
- WizTree. The time is the wall clock for
WizTree64.exe "C:\" /export=out.csv /admin=1 /exportmaxdepth=1. That includes process startup and writing a small CSV. - WinDirStat. The time is the wall clock for
WinDirStat.exe /saveto out.csv C:\. That includes process startup and writing every item to the CSV, which came to 518 MB forC:\. Some of its time goes to that file.
Disk, CPU, file count and the filesystem cache all move these numbers. On this machine, the first directory walk after the other tools ran took 28.6 s. Later walks took 11 to 19 s.
To reproduce the Sprawl numbers, run these from an elevated terminal in
src-tauri:
cargo test --release mft_matches_directory_scan -- --ignored --nocapture
cargo test --release smoke_real_directory -- --ignored --nocapture
The first command scans C:\ through the MFT and then through the directory
walk, and prints both times. Set SPRAWL_MFT_DRIVE to scan another drive. The
second command scans C:\Windows\System32 by default. Set
SPRAWL_SMOKE_PATH=C:\Windows to match the folder benchmark.
- Direct MFT reads. For a whole NTFS drive, Sprawl reads the master file table with overlapping unbuffered reads and builds the tree from its records, without opening a single directory. If anything goes wrong, it falls back to the directory walk.
- Parallel walk. Folder scans and non-admin scans run one Rayon task per directory.
- Compact tree. Nodes sit in one arena and store only their own name, not the full path. Sprawl rebuilds paths from parent links when it needs one.
- Small IPC payloads. The UI asks the backend for only what it draws, three levels and 100 children per folder. Sprawl merges the rest of each folder into one "N smaller items" entry, so totals still add up.
- Fast reopen. Sprawl stores saved scans as zstd-compressed bincode and reopens them without rescanning the disk.
Sprawl is a Windows app. Download the -setup.exe installer from the
latest release.
It asks whether to install for everyone on the computer, which needs
administrator rights and installs to Program Files, or only for you. Either
way it puts sprawl-cli on PATH, the system PATH for everyone or your user
PATH for you alone, and removes it again when you uninstall.
To build from source, you need:
- Node.js 20.19+ or 22.12+
- stable Rust
- the Tauri prerequisites for Windows, which are the Microsoft C++ Build Tools and WebView2
npm install
npm run tauri dev # run in development
npm run tauri build # build an installer under src-tauri/target/release/bundle
npm run dev opens the UI in a plain browser, but it can't scan. Only the
Tauri backend reads the disk.
- Pick a drive on the home screen, or choose Scan a folder….
- Click folders in the treemap to zoom in. Use the breadcrumb or Backspace to go back.
- Right-click an item to open it, reveal it in Explorer, copy its path or delete it.
Search. Type in the search field, or press Ctrl+F. Every term must match, and names ignore case:
| Term | Matches |
|---|---|
report |
names that contain report |
"two words" |
names that contain the phrase |
*.iso, report-??.pdf |
whole names, with * for any text and ? for one character |
size>1GB, size<=500MB |
logical size, with >, <, >= or <= and B, KB, MB, GB or TB |
type:video |
files of a category: video, images, audio, documents, archives, code, apps or other |
kind:file, kind:dir |
files or folders only |
modified<2025-01-01, modified>=2024-06-30 |
whole UTC calendar day; >, <, >= or <= (<= includes that day; > starts the next day) |
age>1y, age<30d, age>=6m |
age relative to now; d = days, w = weeks, m = 30 days, y = 365 days |
The search covers everything under the current folder and follows you as you navigate. Click a folder in the results to open it. Files use their last-write time; folders use their newest descendant file time. Unknown times never match date or age terms. Scans saved before format v5 need a rescan for dates.
Export. Export… saves the current folder and everything under it as CSV or JSON, chosen by the file extension. CSV files are UTF-8 with a header row. When dates are available, CSV includes modified times in ISO 8601 UTC (empty when unknown); JSON uses Unix milliseconds or null. Older scans omit the CSV modified column and use null in JSON.
Admin mode. Without admin rights, Sprawl skips folders it can't read and uses the slower directory walk. Choose Restart as administrator on the home screen, or turn on Always run as administrator in the settings menu behind the cog icon.
Explorer menu. Turn on Scan with Sprawl in the settings menu to add it to Explorer's right-click menu. On Windows 11 it sits under Show more options. If Sprawl is already open, the scan runs in that window.
Saved scans. Sprawl keeps them in its local app data folder. Each path keeps its latest and previous scan, for up to ten paths. Clear saved scans removes them all.
sprawl-cli prints JSON reports, so scripts and AI coding agents can ask what
is filling a disk without opening the app. The Windows installer for the app
includes it and adds the install folder to PATH. To install only the CLI, run
this in PowerShell. It needs no admin rights, installs to
%LOCALAPPDATA%\Programs\sprawl-cli and adds that folder to your user PATH:
irm https://github.com/No-Half-Measures/Sprawl/releases/latest/download/sprawl-cli-install.ps1 | iex
Run the downloaded script with -Uninstall to remove it again. Each GitHub
release also attaches the bare binaries, such as sprawl-cli-windows-x64.exe.
To build it yourself, run cargo build --release --bin sprawl-cli in
src-tauri.
sprawl-cli scan C:\ # scan, save, and report
sprawl-cli show C:\Users\me\AppData # report from the newest saved scan
sprawl-cli find C:\ "*.iso size>1GB" # search the newest saved scan
sprawl-cli find C:\ "size>1GB age>1y" # large files untouched for a year
sprawl-cli export C:\Users --output users.csv # export a saved scan
sprawl-cli list # list saved scans
A report covers the path's size, its largest children (--depth levels deep,
--limit per folder), its largest files, type and extension breakdowns
(top --limit extensions), and changes since the previous scan. It also includes the scan's age, so an agent can tell when
to rescan. Reports include modified times (Unix milliseconds, 0 when unknown)
and scan.hasModified to indicate date availability. show works for any
folder inside a saved scan, so one drive scan answers questions about every folder on it. find takes the same queries as
the app's search and prints the largest --limit matches with full paths and
the total match count. export writes the same CSV or JSON as the app; use
--format or let the --output extension decide. The CLI shares saved scans with
the app and never deletes files. Run it elevated to scan whole drives from the
master file table.
skills/sprawl is an agent skill that tells Claude
Code, Codex and other agents when to use the CLI, how to read its reports and
which files not to delete. Each release also attaches it as sprawl-skill.zip.
Copy the sprawl folder into ~/.claude/skills/ for Claude Code or
~/.agents/skills/ for Codex.
- Windows only.
- The treemap, sorting and percentages use logical file sizes. Size on disk appears beside them. Scans saved by older versions of Sprawl don't have it, so it stays hidden until you rescan.
- The directory walk, used without admin and for folder scans, counts every name of a hard-linked file, both in size and in size on disk. Only MFT scans count a hard-linked file's space once.
- Sprawl doesn't follow symlinks or junctions. Cloud placeholders, such as OneDrive files, count at their logical size, and only their downloaded parts count towards size on disk.
- Results are a snapshot. Rescan to pick up files that changed afterwards.
npm run build
npm test
cd src-tauri
cargo test
cargo clippy --all-targets -- -D warnings
cargo fmt --check
Tests create temporary fixtures. The link test tries a symlink, then a junction,
and reports a skip if Windows allows neither. Tests that scan real drives or
call the Windows shell to delete files carry #[ignore]. Run them with
--ignored.
DESIGN.md has the full product specification, IPC commands and saved-scan format. A few details live only here:
get_treecaps depth at 8 and children per folder at 500, and returns about 8,000 nodes at most. Omitted siblings become one synthetic node withid: null.get_largest_filesandsearchreturn at most 1,000 items.get_trail(nodeId)returns the shallow nodes from the root to an item, so folder search results can be opened.get_type_breakdownreturns all eight categories, including empty ones.start_scanreturns a scan ID, and every scan event carries it asscanId, so the UI can ignore late events from an older scan. The events arescan-progresswith{ scanId, files, folders, bytes, errors, currentPath },scan-completewith{ scanId, root, errors, hasAllocated },scan-cancelledwith{ scanId }andscan-errorwith{ scanId, message }.get_node_pathbacks Copy path, andopen_folderopens a scanned directory.- src/categories.json defines category IDs, labels, colours and extensions. The Rust backend and the React frontend share it.