grab is a media downloader built for the fysh.site ecosystem. paste a link from almost any platform and grab downloads it directly to your device -- no accounts, no trackers, no ads.
grab was built because downloading something you already have access to shouldn't require a dozen browser extensions, sketchy websites, or a tutorial. paste a link. get a file. that's it.
a tool for saving content you want to keep. a clip from a stream, a video essay you want offline, music from soundcloud, a reel you want to watch without an app. grab handles the technical parts so you don't have to.
grab is not a piracy tool. it does not bypass paywalls, crack DRM, or access content you wouldn't otherwise be able to see. it simply makes downloading publicly accessible content easier.
grab is built and maintained by one person as part of the fysh.site project. it runs on cobalt, an open-source media processing engine by imputnet. without cobalt, grab wouldn't exist.
when you paste a link, grab sends it to a self-hosted cobalt processing instance. cobalt resolves the link, figures out what kind of content it is, and returns a direct stream URL or proxied tunnel.
for most content, cobalt returns a direct URL pointing to the platform's own CDN. your browser downloads the file straight from the source. grab's server is not involved in the transfer at all -- it just resolved the URL.
some platforms block direct access from unfamiliar clients. in those cases, cobalt acts as a proxy tunnel. the file passes through the server briefly and is immediately discarded after delivery. nothing is written to disk.
some downloads -- like high-quality YouTube videos -- come as separate video and audio streams that need to be merged. on supported browsers, this merging happens entirely on your device using ffmpeg compiled to WebAssembly. your hardware does the work, not the server.
if you're a Library member and choose to save something to your collection, the download is queued to a separate worker that fetches and stores the file in your personal Jellyfin library. this is the only case where content is written to disk -- intentionally, at your request.
grab supports any platform that cobalt supports. this includes YouTube, Twitter/X, Instagram, TikTok, Reddit, Twitch, Vimeo, SoundCloud, Bilibili, Dailymotion, Facebook, Pinterest, Snapchat, Bluesky, Loom, and Streamable. platform support can change as platforms update their APIs.
grab is designed around minimal data collection. here is exactly what is and isn't collected.
grab does not log what URLs you download. it does not store your IP address in relation to your downloads. it does not use analytics, tracking pixels, or third-party scripts. it does not sell or share any data.
for users without an account, grab tracks only a download count per IP address within a rolling 24-hour window. this count is used to apply fair-use limits and is not stored permanently. it is deleted when the window expires.
if you are signed in as a Library member, grab logs a record of your download history -- the URL, the platform it came from, and the timestamp. this is used to display your recent downloads in the interface. this data is associated with your account and is visible only to you.
you can delete your download history at any time from your Library settings.
content you explicitly save to your Library collection is stored on the fysh.site server and accessible through your Jellyfin account. this is intentional storage at your request. the content counts toward your storage quota and can be deleted by you at any time.
grab's download processing is handled by a self-hosted cobalt instance running on fysh.site infrastructure. your requests do not pass through cobalt's public servers. the cobalt instance operates under the same privacy principles described here.
questions about privacy can be sent to [email protected].
by using grab, you agree to use it responsibly and ethically. these terms are short because the principles are simple.
grab is a tool for downloading content you have legitimate access to. personal archiving, offline viewing, content you created yourself, publicly available media -- these are all fine uses.
you are responsible for what you download and what you do with it. grab provides a technical capability. how you use that capability is on you.
please be mindful of creators. if you download someone's work, credit them when you share it. don't distribute content in ways that harm the people who made it.
grab is not for downloading content you don't have access to. it is not for bypassing paywalls or accessing region-locked content you wouldn't otherwise see. it is not for mass scraping or automated bulk downloading.
grab and fysh.site take no responsibility for how downloaded content is used or distributed. the tool processes requests -- the decisions are yours.
abuse of the service -- including automated scraping, circumventing rate limits, or using grab to download content for harassment -- will result in access being revoked. report abuse to [email protected].
these terms may be updated as grab evolves. significant changes will be noted in the about section.
grab exists because of open-source software and the people who build and maintain it. this page acknowledges what grab is built on.
to the cobalt team for building something genuinely useful and keeping it open. to the yt-dlp contributors for maintaining a tool that refuses to die no matter what platforms throw at it. to everyone who builds open-source software and puts it out there for free -- grab wouldn't exist without you.
grab.fysh.site -- part of the fysh.site ecosystem.
download
cobalt
grab runs on cobalt -- an open-source media downloader built and maintained by imputnet. if grab has saved you time, consider supporting the people who made it possible.
support cobalt on Boosty sponsor cobalt on GitHubfysh.site
grab and the fysh ecosystem are built by one person in their spare time. if you want to say thanks, any support helps keep the servers on.
support fysh.site on Ko-fi