See it before you change it
Adding a folder starts a read-only analysis. Nothing is compressed until you say so.
- Largest files and folders, with drill-down
- Opportunity estimate from sampled Deflate
- Unknown sizes show a dash, never a fake zero
ZipXP shrinks the files in the folders you pick, in place, using Windows' built-in NTFS compression. No archives, no renames, nothing deleted. Each successful change is measured, and Restore original size decompresses them again.
Native to the platform. Nothing extra.
Find out where the space went, decide which folders ZipXP may touch, and let a single, polite background worker handle the rest. You get the measured results.
Adding a folder starts a read-only analysis. Nothing is compressed until you say so.
Turn on automatic compression per folder. New, copied, replaced and changed files are picked up on their own.
Each successful, measurable operation records its signed before-and-after allocation once, in a daily ledger kept apart from estimates.
before − after allocationsThese screenshots come from the actual app, taken after its own engine compressed a generated sample folder of logs, CSV, JSON, source and photos. Your savings will differ.
Add, review, decide. ZipXP keeps doing the rest from the tray while it's running, even with the window closed.
Pick a local fixed NTFS folder like C:\Users\you\Projects. Overlapping roots are rejected up front.
Allocation totals, candidate bytes, largest files, and how many directories already carry Windows' own compression flag.
Turn on automation, choose subfolders and battery behavior. Advanced options hold thresholds, schedule and reserve.
Eligible files join one persistent, sequential queue. File-system notifications and a 2-minute reconciliation scan keep it current.
One file at a time, below-normal priority, through a validated native handle. Busy files retry with backoff.
Actual allocation is read back from the same handle. Poor-benefit files are cached and left alone until their content changes.
Gates are checked between operations, so the worker won't start a new file while a condition says wait. Toggle them below to watch the simulated queue respond.
Turn a condition on to make the worker wait.
Compression is a filesystem attribute, not a rewrite into some other format. ZipXP still treats every operation as risky: it pins the file, proves the file is the same one it scanned, and checks the result. That makes races much less likely, but it isn't a guarantee against every concurrent change, so start with copies.
Open one handle to the original with no write or delete sharing.
Resolved path and stable FILE_ID_INFO must match the scan.
Length and last-write time are re-checked fresh. Changed files wait for a rescan.
FSCTL_SET_COMPRESSION on that same handle. One file at a time.
Actual state and allocation are read from the handle and written to the ledger exactly once.
Some files are excluded from changes outright, and a few kinds of location are refused altogether. ZipXP can't read a file's size or attributes? It shows an error. It doesn't count that file as zero.
File types never changedLocations refused
Most “space saved” counters mix guesses with facts. ZipXP keeps what the drive reports, what it scanned, what it estimates, and what it actually measured in four separate places.
Live drive statistics. Includes activity from every other app, so it moves on its own.
Primary-stream allocation from each root's latest scan, with its timestamp and how much coverage is unknown.
Three 32 KiB Deflate samples per eligible file, scaled up. A heuristic, clearly labeled. Sampling never modifies anything.
Signed difference for successful operations with stable identity. Zero is valid. Decompression is negative.
main yet.Clone it and run it with the .NET 10 SDK, launch it from WSL with the Windows SDK, or publish a self-contained x64 package that needs no separate .NET runtime.
--fresh, --rebuild and --reset flags. Ctrl+C asks the app to quit cleanly.# Install the latest release for this user (no admin) PS> curl.exe -fsSL https://kiranscript.github.io/zipxp/install.ps1 -o "$env:TEMP\zipxp-get.ps1" PS> powershell -NoProfile -ExecutionPolicy Bypass -File "$env:TEMP\zipxp-get.ps1" # Optional: start at sign-in and launch now PS> powershell -NoProfile -ExecutionPolicy Bypass -File "$env:TEMP\zipxp-get.ps1" -StartAtLogin -Launch
# Clone and run (Windows, .NET 10 SDK) PS> git clone https://github.com/kiranscript/zipxp.git PS> cd zipxp PS> dotnet run --project src/ZipXP.App/ZipXP.App.csproj # Run the full native + UI check harness PS> .\scripts\check.ps1
# Launch the native Windows app from WSL $ ./scripts/dev.sh # Throwaway data, your real state is untouched $ ./scripts/dev.sh --fresh # Clean bin/ and obj/ first, then start empty $ ./scripts/dev.sh --rebuild --fresh # Fully isolated instance $ ./scripts/dev.sh --data-dir 'C:\Temp\ZipXP-isolated'
# Self-contained Windows x64 app + ZIP under artifacts\ PS> .\scripts\publish.ps1 # Verify isolated install, launch, exit, uninstall PS> .\scripts\package-check.ps1 # Install for the current user (from the extracted ZIP) PS> .\install.ps1 -StartAtLogin -Launch
ZipXP never deletes, archives, or moves your files. Restore original size decompresses the files it manages when there's room for them, and every recursive action shows you its scope before it runs. Windows' own per-folder compression flags are a separate switch.
No. ZipXP uses Windows' built-in NTFS compression. Files keep their original names, paths and extensions, and every app opens them exactly as before. Windows decompresses them on the fly when they're read.
It depends entirely on what's in the folder. Text, source code, documents and logs shrink well. Photos, video and archives are already compressed and barely change, and ZipXP skips files that don't benefit.
The in-app estimate is a heuristic from Deflate samples, but NTFS uses a different algorithm, so treat the estimate as a hint. The measured number is what counts.
ZipXP runs one below-normal-priority process with a single sequential worker, a 500 ms interval between files, and pauses new operations when sampled CPU is above 85%. These gates are checked between operations. They aren't hard resource caps.
Reading compressed files costs some CPU. How much depends on your files, hardware and workload, and broad performance measurements are still to come, so try it on representative tasks first.
No. It's a per-user app with a tray icon. Signing in only starts it if you explicitly turn on Start when I sign in.
Jobs that were running become retry records, and the file's actual state and identity are checked again. If an interrupted operation had already compressed a file, ZipXP records it as unchanged instead of inventing a saving. Files changed while ZipXP was closed are found on the next startup.
Not yet. ZipXP 0.2 is an unsigned pre-release. The development checks pass, but broader hardware and antivirus compatibility testing, a release signing certificate, and a real user pilot are still to come. Use disposable copies of your data until then. See boundaries.
Yes, once a release is published: a one-line curl.exe command downloads the ZIP, verifies its SHA-256 and installs per user (see the quickstart). Release ZIPs carry a GitHub build attestation, and the script checks it when the GitHub CLI is installed. The build is not code-signed. You can also build it from source with the .NET 10 SDK, or run scripts\publish.ps1 to make the same self-contained ZIP yourself.
Policies, observations and history live locally in %LOCALAPPDATA%\ZipXP\StorageManager\storage.sqlite3. Nothing is uploaded, and there's no telemetry.
Clone the repo, start it with dotnet run, and add a folder of copies. ZipXP analyzes it before it changes a single byte.