01 · Privacy
Private by architecture.
Your file is read and transformed by code on your device. There is no file-processing server to trust.
Private by architecture
Convert, organize, and clean files in your browser. No account, upload queue, or server copy.
Compress PDF
ExampleReading bytes0.0 / 2.4 MB
network requests since this loaded0
87 focused utilities
Search by task or format. Press / or Ctrl K.
Search all 87 tools.
No tools match that search. Try a format such as PDF, CSV, JPG, or JSON.
Welcome back
Common tasks
Fast paths for everyday file work.
Everything there is
Image
Data & CSV
Text
Developer
Or the whole catalogue
Built differently
01 · Privacy
Your file is read and transformed by code on your device. There is no file-processing server to trust.
02 · Access
Open a tool, finish the task, and download the result. Core tools are free and require no identity.
03 · Limits
We do not meter file uploads because files are not uploaded. Your device and each format set the practical boundary.
04 · Offline
After a tool and its processing assets load once, FileGizmo caches them on this device so the same workflow can run without a connection.
Two things worth knowing
Most sites ask you to believe a retention policy. This one makes a claim a browser can check: open the network panel, run a tool, and count the requests that carry your file. The answer is none, because there is no endpoint to send one to.
Rather than leave that as an assertion, the proof page runs a real conversion in front of you and reports the bytes in, the bytes out, the time taken, and the number of off-origin requests. A build stage fails if that last number is ever anything but zero.
Watch it runFileGizmo ships a Model Context Protocol server, so an assistant such as Claude or Cursor can decode a JWT, merge record sets, or read PDF page ranges as a step in its own work rather than sending you to a website to do it by hand.
It runs as a local subprocess and imports the same modules the site and its tests run, so an answer there is the answer you would get in the browser. Read-only, nothing uploaded, no key to obtain.
Set it upStraight answers
No. Every tool runs inside your browser, using code your device downloads once and then executes locally. There is no upload endpoint, so there is nothing to send a file to. You do not have to take that on trust: open your browser's network panel, run any tool, and watch the request list stay empty. A build stage fails if a page ever contacts a third-party host, and /proof/ replays the measurement live.
It is gone. The file is held in the tab's memory while you work and released when the tab closes. Nothing is written to a server, because there is no server in the path, and nothing persists on our side to delete later — which also means we cannot recover a result for you, so download it before you leave.
No. No tool on this site asks for an email address, a sign-in or a verification step, and none of them will. There is no free trial that expires, no daily quota, and no watermark on anything you download.
Every tool listed on this page is free to use with no account. The honest catch is that the model is different rather than absent: because files never reach a server, there are no per-file processing costs to recover, which is what makes free sustainable here. There are no advertisements on any tool page, and there never will be.
None that we impose, because we are not paying to store or move your file. The practical ceiling is your own device: how much memory the browser tab can use for the file and the working copy it makes. Large PDFs and high-resolution images are the usual limits, and /upload-limit-checker will tell you what a given file would run into.
Yes, once a tool has loaded. A service worker keeps the page and its processing code on your device, so you can open a tool while connected, disconnect, reload it, and keep working. The first visit needs a connection, and tools that carry a large processing engine need that engine fetched once too. Clearing site data removes the cached copy.
Any current version of Chrome, Edge, Firefox or Safari, on desktop or mobile. The tools use standard browser capabilities — Web Workers, WebAssembly and the Canvas and File APIs — rather than anything vendor-specific. JavaScript has to be enabled, since the processing is the JavaScript.
Open /pdf/compress, choose your file, and the work happens on your device. Embedded photographs are re-encoded, text stays selectable and vector art stays sharp. If a file is already as small as it can get, the tool says so rather than handing back a copy and a zero — and if you need a specific size, /pdf/compress-to-200kb and its siblings are the same tool with the target filled in.
Yes. FileGizmo publishes an MCP server exposing 42 of these tools to assistants that speak the Model Context Protocol, so an agent can merge, split, convert or hash a file as a step in its own work. /mcp/ has the connection details and the full tool list. No competing file-tool site offers this.
Most of them upload your file, process it on their servers and delete it afterwards on a schedule you have to believe in. That is a reasonable design, and it is a different one: it means your document leaves your machine. FileGizmo does the work where the file already is. /vs/ compares the specifics against the sites people usually ask about, including where those sites are genuinely better.
No ceremony