Kadr — screenshots & step guides

Capture any page, beautify the shot and turn clicks into a step guide. No trial, no watermarks, no PRO lock.

Take the shot. The whole page stitched from frames, the visible area, a region you drag with a pixel magnifier, or a DOM element picked by hovering it. A chat or an endless feed is captured while it scrolls, stitched by content rather than coordinates, so inner scroll containers work too. Hover a table and copy it as CSV, Markdown or JSON instead of a picture. A responsive series shoots 375, 768 and 1440 px in one go and puts all three on one canvas. A saved shot can be reshot from the same page at the same width with one click. Make it presentable. Backgrounds drawn in code — gradients, patterns, solids — margins, corner radius, shadows, real macOS and Windows 11 browser frames with an editable url, device bodies, and a mockup of your own. Save the whole look as a style preset and share the JSON with your team. Explain it. Arrows in seven styles, numbered step badges, spotlight, highlighter, text with your local fonts, shapes and callouts. Blur stays a layer you can move or remove later instead of pixels burnt into the image. Turn clicks into a guide. Press Record, walk through the interface, press Stop. Every click becomes a step: a frame with the element outlined, a numbered badge and a caption like "Click 'Save'" taken from the element's own label. Restyle the whole guide at once, redact private data across every step offline, and export it as Markdown with images, a PDF or one long image. Hide what should not be public. Text recognition runs inside the browser — no key, no server, no upload — and finds emails, phone numbers, cards and tokens for you to cover with one click. Nothing is redacted without your confirmation, and nothing is ever a guarantee: check the shot before you publish it. Send it where the work is. A library with search by site, date, tag and the text recognised on the shot. Copy as an image, as markdown or as a data URI. Export PNG, JPEG, WebP or PDF. Or go straight to a GitHub, Linear or Jira issue with your own token — the picture, the page url, the browser and the screen size come along. Bring your own AI, or none at all. AI is off by default and makes no request until you turn it on and enter a key. Write your own prompts and run them on the shot, or edit the picture with words and get the result back as a layer. OpenAI, Anthropic, Google, a local Ollama or LM Studio, or any OpenAI-compatible endpoint. The redaction pass above needs no AI at all. No telemetry. Not a line about you goes anywhere: no counters, no analytics, no "anonymous statistics". The extension makes exactly three kinds of network request, all of them started by you: an AI action with your key, a one-time dictionary download for offline text recognition, and sending a shot to the tracker you configured. Open source, MIT: https://github.com/crocone/kadr