Extract tar.gz Files Online

Extract tar.gz files online: one tarball up to 50 MB — .tar, .tar.gz, .tgz, .tar.bz2 or .tar.xz — is unpacked on your own device. Split .tar.001 sets, .tar.zst and single-file .gz are not supported.

Drop files here or choose files
Your tarball stays on your device · unpacked in the browser · tar, tar.gz, tar.bz2, tar.xz · one archive up to 50 MB

How to use

  1. Choose a tarball: Drop a .tar, .tar.gz, .tgz, .tar.bz2 or .tar.xz file onto the box, or click to choose one. One archive at a time, up to 50 MB; to swap it for another, press ✕ on the file row first.
  2. Press Extract TAR: The first run fetches the extraction engine — a 60 KB worker and a 0.96 MiB WebAssembly file — so the progress bar waits at 0% for a moment; after that it counts the entries as they come out. Cancel requests cancellation, and no partial result is ever returned.
  3. Preview the files and save them: Every file is listed with its full path inside the archive and its extracted size. Use the numbered buttons to look at one and Download it, or Download all to save them one after another.

How to extract a tar.gz file online without installing anything

Pick one tarball of up to 50 MB — .tar, .tar.gz, .tgz, .tar.bz2 or .tar.xz — press Extract TAR, and every file inside is listed with its path and its size, ready to preview and save. The archive is unpacked on your own device, inside the browser tab, and is not uploaded to any server.

The decoding is done by libarchive compiled to WebAssembly, running in a worker thread so the page keeps responding while the entries come out. Both the engine and the tar code load on demand, which means opening this page downloads no decoder at all: the first press of the button fetches a worker of about 60 KB and a 0.96 MiB WebAssembly decoder from this site, and that engine is the only thing that travels over the network. The RAR extractor here runs on the same decoder, so if you unpacked a .rar first it is already in your browser cache.

The file is identified by its bytes, not by its name. Three outer signatures are read at offset 0 — 1F 8B for gzip, BZh for bzip2, FD 37 7A 58 5A 00 for xz — and an uncompressed tarball is recognised by the ustar marker that sits at offset 257. Older v7 tarballs carry no marker at all, so the first 512-byte header block is accepted when its stored checksum matches the bytes around it. If nothing matches, the file is reported as damaged there and then, before the WebAssembly decoder is downloaded, so a .zip that somebody renamed to .tar.gz does not cost you a megabyte of traffic first.

Nothing is kept anywhere. The archive is read straight from your disk into the tab's memory, the extracted files stay there, and the previews and download links point at those in-memory copies. There is no cookie and no local storage: a refresh or a closed tab clears all of it, and only the files you saved yourself remain on your device.

Which tar formats can I extract online?

Five extensions, one code path. libarchive reads the tar container through the outer gzip, bzip2 or xz filter transparently, so a compressed tarball needs no separate step and you never have to tell the page which flavour you have — .tgz, .tbz2 and .txz are simply short spellings of .tar.gz, .tar.bz2 and .tar.xz, and they are handled identically.

An archive that is genuinely empty is accepted rather than called a failure: a tar ends with two 512-byte blocks of zeros, and a file that contains only those is a valid empty archive, so the page says the archive is empty instead of reporting an error. That ruling applies to uncompressed tarballs — a compressed file with no entries is a different case, covered in the next section.

The notable gap is Zstandard, which is not compiled into this libarchive build: .tar.zst is not supported here. The file picker may reject it; if it reaches extraction, it is reported as damaged or not a valid tar archive. Those archives need tar or 7-Zip on a computer.

It helps to know what the two halves of the name do, because it explains the sizes you will see. tar only concatenates files into one stream with a 512-byte header in front of each; the compression is the outer layer added afterwards. That is why an uncompressed .tar is normally larger than its .tar.gz, and why the difference is widest for archives full of small files, which are mostly headers and padding. The sample archive used in this tool's tests holds four files with 125 bytes of content between them: as a .tar it is 10,240 bytes, as a .tar.gz it is 273 bytes. That ratio belongs to that particular sample and is not a rule for tarballs in general.

FormatCompressionSupported
.tarnone, packaging onlyYes
.tar.gz · .tgzgzipYes
.tar.bz2 · .tbz2bzip2Yes
.tar.xz · .txzxzYes
.tar.zstzstdNo — zstd is not in this libarchive build
Single-file .gz, .bz2, .xzcompression with no tar insideNo — not a tar archive

What this page reads, decided by the bytes in the file rather than by the extension.

If the archive you were sent is a .rar rather than a tarball, Extract RAR Files Online does the same job on the same engine.

Why a single .gz, .bz2 or .xz file is turned away

A plain compressed file is not a tar archive, and this tool refuses it on purpose. access.log.gz and project.tar.gz begin with the same two bytes, so the outer signature can only prove that the stream is gzip — it says nothing about whether a tar archive is hiding inside. The same is true of bzip2 and xz streams.

The giveaway comes after the stream is opened: libarchive unwraps a single-file gzip without complaining and then reports no entries at all. A zero-entry result from a compressed file is therefore ruled a single compressed file rather than being shown to you as an empty archive, and the message names what you actually have. If you need the contents of a lone .gz, decompress it with gzip, 7-Zip or Windows' built-in tar command instead.

Split sets are recognised by name and refused before the extraction engine is downloaded. The pattern covers name.tar.001, name.tar.gz.001, name.tar.bz2.001 and name.tar.xz.001 — three digits, upper or lower case — so a file called report.tar.01 is not mistaken for a volume. The reason is structural: a browser receives only the single file you picked, and the decoder needs one complete buffer it can seek around in, so there is nothing for it to join. Put the parts together on a computer first, then bring the whole archive here.

The drop box itself accepts .tar, .gz, .tgz, .bz2, .tbz2, .xz and .txz, which is why a single-file .gz gets as far as the button before the format check turns it away. Other file types are filtered out at the drop box, and a file over 50 MB is stopped there with a size notice.

If what you have is a .zip — the format a tarball is usually repacked into for Windows — use Unzip Files.

What comes out of the tarball, and what it loses

File contents come back byte for byte. Extraction decodes, it does not convert: there is no re-encoding step and nothing is compressed a second time, so a source file, a photo or a database dump is bit-identical to what went into the archive. The tarball itself is only ever read — it is not rewritten, repaired or repacked, and it stays exactly as it was on your device.

The list shows each entry's full path inside the archive, such as docs/notes.md, next to its extracted size, so the original folder structure stays visible at a glance. Modern tar stores names in UTF-8, and ASCII names and names using characters beyond Latin-1 — Chinese characters and emoji included — come through as they are; names written entirely in Latin-1 characters carry one caveat, below.

Download names are a different matter, because a browser saves single files and will not rebuild folders. Folder separators are folded into underscores, so docs/notes.md arrives as docs_notes.md. Different paths can produce the same download name: a/x.txt and a_x.txt both become a_x.txt. Repeated archive paths also produce duplicate names. Check saved files and rename them when necessary — the list on the page keeps the full paths, so you can tell which entry each download came from.

Several things recorded in a tar header are not carried anywhere. Permission bits, owner and group, and modification times are not passed to the page and cannot be written into a browser download, so the files you save carry the time you saved them. Empty folders have no file to carry them and do not appear in the list. Symbolic links, hard links and device nodes are not extracted either — only regular files are, which is worth knowing before you use a browser extraction to move a Linux backup around.

One name-decoding caveat is worth stating plainly: paths pass through a recovery step written for old archives that stored names in the Chinese GBK code page. A name made up entirely of Latin-1 characters that also happens to be a valid GBK sequence can be re-read as Chinese characters, both in the list and in the download name. The file's contents are untouched, so rename it after saving if that happens, or extract that archive with desktop software when the names have to survive exactly.

Limits, cancelling, and when another tool fits better

Archive: 50 MB
Entries: 10,000
Single extracted file: 256 MB
Total extracted size: 512 MB

One archive at a time, up to 50 MB, and only the tar family. A second pick replaces the first rather than queueing behind it, and the ✕ button on the file row is disabled while an extraction is running.

The three content ceilings are checked before a single byte is expanded. Listing the archive gives the size each entry declares, and the whole archive is refused if it crosses one of them — more than 10,000 entries, a single file over 256 MB extracted, or more than 512 MB extracted in total — so a small file that claims to expand into more than your device can hold is stopped at the listing. The running total is checked again against the real bytes as the entries come out. An entry whose assembled path points outside the extraction folder rejects the archive as well: a path starting with / or a drive letter, a UNC path, or any name with a segment equal to exactly .. — an ordinary report..txt is fine.

The whole archive is extracted in one pass. There is no way to list the contents first and take only the two entries you want, and that also sets the memory bill: peak usage is roughly the size of the archive plus everything extracted from it, and that pressure belongs to the whole tab. The decoding runs in a worker, so the page itself keeps responding, but an archive near the 50 MB limit can be heavy going on an older phone.

Cancel requests cancellation rather than guaranteeing an instant stop. Once the entries are being extracted, the stop takes effect between them; while the engine is still loading or the archive is being opened, the current step may have to finish before the run reports that it was cancelled. Either way a cancelled run returns no partial result and never turns back into a successful one.

The progress figure counts entries, not bytes: it is the number of files already extracted divided by the total number in the archive. An archive holding one large file and a handful of small ones therefore sits on the same percentage for a long time and then finishes in a rush, which is normal and not a sign that anything has stalled. On the first run the engine download sits in front of all of it, which is why the bar waits at 0% to begin with.

This page never writes a tar. It only reads one, so there is no way to build a .tar.gz here, and there is no password field either: the tar format has no encryption of its own, so a password-protected archive you were sent is a .rar, .7z or .zip rather than a tarball.

To bundle the files you just extracted and pass them on, use Create ZIP — a .zip opens on Windows, macOS, Linux, Android and iOS without extra software.

Frequently asked questions

Is my tar.gz file uploaded to a server?
No. The archive is read from your disk and decoded inside the browser tab, and the previews and download links point at copies in your browser's memory. The one thing that does come over the network is the extraction engine — a 60 KB worker and a 0.96 MiB WebAssembly file, served from this site — and it is fetched the first time you press the button, not when the page opens.
Does it open .tar.bz2 and .tar.xz as well as .tar.gz?
Yes, and .tar, .tgz, .tbz2 and .txz too. The decoder reads the tar container through the outer gzip, bzip2 or xz layer, so all of them take the same path and need no setting from you. The exception is Zstandard: .tar.zst is not supported here. The file picker may reject it; if it reaches extraction, it is reported as damaged or not a valid tar archive.
Can I open a single .gz file like access.log.gz here?
No — that is a compressed file, not a tar archive, and the tool says so instead of failing vaguely. A lone .gz and a .tar.gz start with the same two bytes, so the refusal comes after the stream is opened and turns out to hold no archive entries. Decompress it with gzip, 7-Zip, or the tar command built into Windows 10 and 11 instead.
Why won't it extract my file.tar.gz.001 archive?
Because that file is one volume of a split set, and a browser only receives the single file you picked — there is nothing to join it to. Names ending .tar.001, .tar.gz.001, .tar.bz2.001 and .tar.xz.001 are recognised and refused with a clear message rather than half an archive. Join the parts on a computer, then bring the complete tarball back here.
How big a tarball can I extract?
Up to 50 MB per archive, one at a time. The contents are capped as well: 10,000 entries, 256 MB for any single extracted file, and 512 MB extracted in total. An archive that declares more than any of those is refused at the listing stage, before anything is expanded, so it cannot quietly fill your device's memory.
Why do the folders disappear when I download the files?
A browser can only save single files, so the folders are not rebuilt: docs/notes.md arrives as docs_notes.md. Different paths can produce the same download name — a/x.txt and a_x.txt both become a_x.txt — and an archive that stores the same path twice gives both copies the same name as well, so one can overwrite the other. Check the saved files and rename them when necessary; the full paths stay in the list on the page, so you can recreate the structure yourself.
Does extracting change the files or their dates?
The contents are unchanged, byte for byte, because extraction only decodes. The metadata is a different story: permissions, owner, group and modification times live in the tar header and cannot be written into a browser download, so saved files carry the time you saved them. Symbolic links, hard links and device nodes are not extracted at all.
How do I open a tar.gz file on Windows without installing software?
Choose one .tar.gz file up to 50 MB on this page and press Extract TAR. Preview and download the extracted files in your browser. Downloads flatten folder paths and do not preserve Unix permissions or original modification times.
How do I open a tar.gz file on an iPhone?
Open this page in Safari or Chrome, tap the drop box, choose the archive from Files or iCloud Drive, and press Extract TAR. The iOS Files app unzips .zip but not .tar.gz, which is why the usual advice is to install an archiver app. Save the files you need one at a time; Download all saves them in sequence, which on a phone can mean one prompt per file.