Free
Everything local, forever.
- Record, annotate, redact, export
- Captures stay in your browser (OPFS)
- A handful of Distill runs a day
- Unlimited AI against your own local model
- The Chrome extension and all its context
A screenshot shows what broke. A Shotyx capture also carries the click path that got there, the console error underneath, the network call that failed, and the labels of the fields involved — in one file.
The web app records your screen and is the editor. The Chrome extension is what adds the context.
This is a genuine .shotyx bundle, not a video of one. It
imports into your browser’s private storage and opens in the same
editor you’d use for your own work: four screens, seventeen
events on one clock, four drafted steps, and a redaction over the
payment fields you can toggle, move, and export.
Imported locally, deleted whenever you want — like every other capture on this machine.
Screens, clicks, keys, navigations, console output and network calls are recorded against the same timeline, so a step and the error it caused sit next to each other instead of in two tools. Field labels are captured; field values never are.
The event rail below is the demo capture’s, unedited.
Drawing a redaction doesn’t lay a black box over the image — the pixels underneath are resampled away before anything is written, exported, or shared. There is nothing left to peel off.
The rest follows from where captures live: your browser’s origin-private storage, on your own disk. Recording, editing and export all run on-device, and no capture is uploaded unless you deliberately share it.
Distill condenses a long recording into one full-resolution keyframe per action, drafts a sentence for each from the events that caused it, and numbers them across the whole capture. Edit the wording, then export the walkthrough as a numbered document.
These four are the demo capture’s drafted steps.
Point the editor at Ollama or LM Studio on your own machine and the
summary, the reproduction steps and the drafted bug report are written
there. Your browser talks to localhost directly — the
capture never leaves the machine it was recorded on, and there is no
account, no credits and no API key in the way.
Running a vision model? It reads the frames as well as the events. Running a text-only one? It works from the click path, the console error and the failed call, which is most of what a bug report needs anyway.
You install the runtime yourself — ollama.com or lmstudio.ai, a few minutes once. If you’d rather set up nothing at all, server credits draft the same things on our machines instead.
The web app records your screen — that alone is a good bug report. The Chrome extension records what the page was doing while you did it.
Every click with its selector and the label the user actually saw.
Warnings and errors with their stack, on the same clock as the frames.
Method, URL, status and duration. Never response bodies.
Scroll-and-stitch capture with section markers for tall pages.
Coming to the Chrome Web Store Until then, start in the browser — captures move over intact.
Everything local, forever.
For work that leaves your machine.
Capture, editing, export and local AI never need an account. Only the parts that use our servers — shares, hosted Distill runs, AI on our credits — are metered.