September 8th, 2026
Improved
The image and carousel editor got a new canvas. Everything you could do before still works the same way, but a lot of it now happens right on the design instead of in the side panel. Here's what changed and how to use it.
Edit text in place, one word at a time
Double-click any text and start typing where you clicked. Select a few words and the toolbar switches to Formatting selected words: bold, italic, color, size and the new highlighter apply to that selection only. So "Save 50%" with just the number in orange is a two-second job now, and it exports exactly like that. Cmd/Ctrl + B and I work on the selection too. The box keeps its size and its handles while you type, and only grows when you add a line.
Crop, swap and round images from the toolbar
Select an image and you get Swap, Crop and Corners above it. Crop dims what's outside the frame, shows a thirds grid, and lets you drag the photo around inside it. Done keeps it, Reset brings the whole image back.
Group, lock, layers
Select two or more elements and press Cmd/Ctrl + G to group them; from then on they move, resize and delete together (Cmd/Ctrl + Shift + G ungroups). Lock anything that's placed and shouldn't move. The new Layers button at the bottom lists every element top to bottom: click to select, drag to reorder.
Shapes with labels, backgrounds in one click
The add bar at the top drops a text (or press T), an image, a rectangle / ellipse / pill, or a background onto the current slide. Shapes have an Add label button that puts a readable text in the middle and keeps it grouped with the shape. Click the background itself and you get your brand colors and gradients as swatches.
Ask Krumzi about what you selected
Every toolbar ends with a sparkle. Click it and a small menu opens: a field for your own request at the top, and underneath, quick actions that fit the selection, each with its own icon (Rewrite, Shorter, Punchier or Fix grammar on text; Regenerate, Match brand colors or Remove background on an image; Suggest a color on a shape; Harmonize colors when several things are selected). Click one, or use the arrow keys and Enter. Start typing and your own request moves to the top of the list, so Enter sends that instead. The toolbar shows "Krumzi is working on it" while the agent edits, and the toast at the end has an Undo that discards that whole turn.
Shapes, QR codes and the background menu
The add bar's Shape button opens the full picker: Blob, Wave, Rectangle, Circle, Star and Pill, a grid of variants, and a Regenerate roll. A new QR code button takes a link or text. The background's More menu now puts your brand colors, gradients and meshes first, bigger, with the presets below.
Right-click menu
Right-click any element for the quick menu: Edit text, Crop, Fit or Fill the slide, Add label, Duplicate, Copy, Paste, Delete, layer order, Align, Lock, Group, and the full Design panel. Empty canvas gets Paste, Select all, Add text, Upload an image and Edit background.
See the whole carousel at once
Carousels now open with every slide side by side, edge to edge, the way someone will actually swipe through them. A headline that runs across two slides looks like it will in the export instead of getting cut off at the edge of a window.
Click any slide to start working on it. The one you are on has its number in purple, and the toolbar, the layers card and the background all follow it. Move along the strip with the arrows or the dots in the slides pill, or with Page Up and Page Down. Fit (Cmd/Ctrl + 0) frames the whole carousel, and the percentage button in the zoom pill drops you to 100% centered on the slide you are on.
Prefer one slide at a time? The button on the left of the slides pill switches between the two, and so does clicking "2 of 5".
Smaller things you'll notice
Scroll to pan, Cmd/Ctrl + scroll or pinch to zoom at the cursor, Space + drag to move around. A zoom pill at the bottom has Fit and Cmd/Ctrl + 0 fits too.
Orange smart guides snap to slide edges, centers and other elements. Hold Alt/Option to drag without snapping, Shift to lock an axis.
Multi-select with Shift + click or a marquee; resize the whole selection at once; Align and Distribute live in the toolbar.
Every gesture is one undo step and the undo button says what it will undo. Deleting shows an Undo toast.
The slides pill also adds, duplicates and deletes slides, and a single image gets a "Make it a carousel" button.
The AI learned all of it too: ask it to group elements, lock the logo, color one word, or crop a photo, and it uses the same controls.
The side panel is still there for the deep controls, under More → All options or from the left navigation. And if anything feels off, the old canvas is one setting away on our side, so tell us.
September 3rd, 2026
New
You can now connect Krumzi to your AI assistant and design without leaving the chat.
Add the Krumzi server to Claude, ChatGPT, Cursor, or Claude Code, and describe what you want the same way you would in the app: "Make an Instagram post announcing our fall menu, warm and a little playful, use our brand." The assistant starts the design, tells you when it's ready (usually a minute or two), and the finished post is in your Krumzi library. On brand, fully editable, ready to export.
Here's how to try it:
Copy the server URL:
https://www.krumzi.com/api/mcpIn Claude, go to Settings → Connectors → Add custom connector, name it Krumzi, and paste the link. (ChatGPT, Cursor, and Claude Code work too — the help article has steps for each.)
Click Connect and approve on the Krumzi screen that opens.
That's it. Your assistant can create posts, carousels, and documents, edit them, export PNG, JPG, WebP or PDF files, generate AI images, and read your brand kits. If you have more than one brand, just tell it which one to design in — it defaults to the brand you have active in the app.
A few honest notes: designs made this way spend the same credits as designing in the app. Video designs can't come through the connection yet. And the connection can never publish to your social accounts, touch billing, or delete anything.
You manage connections under Settings → Connected apps, where one click disconnects a tool instantly. There's also a page with a live walkthrough at krumzi.com/mcp.
September 1st, 2026
Improved
Visuals got a full rebuild. Same library, much faster to move around in, and it now shows you a real preview of every design instead of a tiny thumbnail.
Search your whole library
The search box in the toolbar looks everywhere, not just the folder you're standing in. Start typing and you'll get matching folders and designs together. Clear it to go back to where you were.
Filter without reloading
Under the toolbar there's a row of chips. Pick a type (All, Images, Videos, Carousels, Docs) or click any of your tags. Everything filters instantly. A dashed Clear chip shows up whenever something is narrowing the view.
Select the way you'd expect
Click to select one. Cmd/Ctrl + click to add or remove. Shift + click to grab a run. Drag a box across empty space to lasso a group. Cmd/Ctrl + A takes everything, Esc clears it.
Drag onto folders, and undo it
Drag designs and folders onto any folder card, or onto a step in the breadcrumb trail to send them up a level. If you grab something while several items are selected, the whole selection comes along. Changed your mind? Every move now leaves a toast with an Undo button.
Right-click anything
Right-click a design or folder for Open, Quick preview, Rename, Duplicate, Move to, Tags and Delete. Right-click empty space for New folder and Select all.
Make tags on the spot
Hit New tag from the filter row, from the Tags popover in the bottom bar, or from the right-click Tags menu. Name it, pick a color, and the preview shows you exactly how it'll look. Anything you had selected gets tagged the moment you create it.
Rename in place
Press F2 (or right-click and pick Rename) to rename right in the grid, folders included. Creating a new folder drops you straight into the name field, so just type and hit Enter.
Peek with the spacebar
Select a design and press Space for a full-screen preview. Use the arrow keys to flick through the rest, Space or Esc to close.
One thing worth knowing: deleting is still permanent, so it asks you to confirm rather than offering an undo. Moving is the reversible one.
August 30th, 2026
New
Brand Mind just got its biggest upgrade yet. Instead of only storing your colors and fonts, it now forms real conclusions about your brand, learns from everything you do, and proves it in your very first session.
Start with just your name. Create a brand and type its name. Krumzi finds your website and logo, pulls your colors and fonts into your brand kit automatically, and gets to work.
Research that does your homework. Krumzi reads your website, looks up what your customers say, and checks how similar brands present themselves, narrating every step while it works. About a minute later you get "Here's what I understand": plain-language conclusions like "You're the premium option, your designs should never look like a discount flyer." Confirm them, fix them in place, or dismiss them. Every design you generate from then on is steered by them.
It speaks your customers' language. If your audience posts in French (or German, or Spanish…), Brand Mind notices, and writes your conclusions and your design copy in that language from the first design on.
No brand yet? Build one together. Pick the interview instead: a real conversation, not a form. Every question builds on your last answer. Tell it you restore mid-century furniture and the next question is about estate sales, not a generic "who's your audience?". Along the way it picks up your brand's name, what you sell, and your colors: if you have colors, say them; if you don't, it proposes a palette from your taste picks. Either way your brand kit is ready by the time the interview ends.
Feed it anything. Drop brand guidelines, a pitch deck, or screenshots of posts you love into Brand Mind. It reads them, shows you what it learned, and asks you to settle it when a document contradicts something it believed.
Watch it learn. The "What I've learned lately" feed shows every signal: a thumbs-down you gave, a font you changed, a design you exported. Anything it learned wrong, undo with one tap.
Open Brand Mind from the sidebar any time to see your brief. The more you react to designs, the sharper it gets.
August 28th, 2026
Improved
We tightened up how the AI designer works behind the scenes, and the result is simple: the same design now costs you fewer credits.
Here's what changed:
The AI checks its work in one pass now. It used to re-screenshot and re-check the canvas over and over while polishing small details, and every extra round cost credits. Now it gathers all the issues first, fixes them in one batch, and takes one final look before handing the design back to you. There are hard limits behind this, so a design can't quietly rack up checks in the background anymore.
Long conversations got cheaper too. In very long design chats, old messages could pile up and make every new request more expensive. Krumzi now trims the oldest parts of the conversation automatically before each request. Your current design is never affected, since the AI always works from the live canvas.
You don't need to do anything to get this. It applies to every design chat automatically.
One small tip: if you're deep into a very long conversation and the AI seems to have forgotten something you said near the start, just mention it again in your next message, or start a fresh chat on the same design.
August 16th, 2026
New
You can now make multi-page documents in Krumzi. Front-and-back flyers, 4-page brochures, menus, one-pager sets, that kind of thing.
How it works
Open the AI Designer and pick Document as the type. It sits right next to Image, Carousel, and Video.
Choose a page size and orientation: A4, A5, A3, US Letter, Half Letter, or DL, portrait or landscape.
Set how many pages you want, anywhere from 2 to 12.
Describe the document like you'd describe any other design. Something like "a 4-page brochure for my yoga studio: cover, class schedule, pricing, contact page" works great.
The AI knows it's designing for print, so it builds a cover, keeps the inside pages on a shared grid, and puts your call to action and contact details on the back. Text stays inside the safe margin, and there are no social-media hooks or platform safe zones getting in the way.
In the editor
Every page shows a dashed safe-margin guide so you can see what's print-safe at a glance. You can add or remove pages, move things around, or just ask the AI to change something ("make page 3 less busy", "swap the cover photo"). Everything's editable, same as always.
Exporting
Hit export and you'll get a multi-page PDF at 300 dpi by default, ready to send to a printer or attach to an email.
Good to know
Documents are available on Pro and Ultra.
June 10th, 2026
New
We just added Claude Fable 5 to the model roster. It's Anthropic's most powerful model yet, and it sits a tier above Opus in the selector.
Reach for it when the brief has to land perfectly on the first try: intricate multi-element campaigns, the sharpest copy, long carousels with real narrative arcs, anything where you want the most capable model available doing the work.
Fable 5 runs at roughly 2x the credits of Opus per turn. It's the priciest option in the lineup, so save it for the moments that genuinely need the best. For everyday work, Sonnet or GPT 5.4 are still the sweet spot.
Available on every plan.
June 7th, 2026
Exporting images and carousels just got an upgrade.
PDF exports. Download any image or carousel as a PDF. Carousels export as a multi-page PDF, one slide per page, ready to share or print.
More formats and sizes. Choose PNG, JPG, PDF, or WebP, at 1× up to 4×. Turn on transparent background for PNG and WebP when you need a clean cutout.
A real export panel. Settings, progress, and download now live in one clean panel anchored to the Export button. Close it mid-export and it keeps running — reopen any time to grab your file.
Better carousels. Export your slides as separate images in a .zip or as a multi-page PDF, and pick exactly which slides to include — each with a live preview.
Previous exports. Every export is saved. Re-download any of your design's past exports right from the panel.
Plus smart file names that auto-version (v2, v3…) so you never overwrite an earlier export.
May 27th, 2026
New
Now you're in control of every edit the AI makes.
When you ask Krumzi to change a design, it no longer just commits the work and hopes you like it. Every element the AI adds, tweaks, or removes shows up for review and you decide what stays.
Here's how it works:
Spot changes at a glance. Each changed element gets a small dot on the canvas, so you can see exactly what the AI touched.
Keep or discard one at a time. Click an element and you'll find Keep and Discard right in its toolbar. Loved that new headline but not the shape it added? Keep one, toss the other.
Or handle it all at once. The review panel in the chat lists every pending change. Hit Keep all or Discard all when you're ready.
See the clean version anytime. Toggle the markers off to preview your design exactly as it'll look, with no review clutter in the way.
Your pending changes stick around too - refresh the page or come back later and they'll still be waiting for your call. And you can keep chatting: the AI builds right on top of changes you haven't reviewed yet, so nothing slows down.
The result: experiment freely. Ask for bold changes, and keep only the ones that land.
May 10th, 2026
New
You can now make Krumzi think before it designs. Flip the new Think pill in the AI chat input and the model takes a planning step before responding. Useful for complex multi-element designs where rushing produces messy results.
Available on Sonnet 4.6, Opus 4.7, GPT 5.4 and 5.4 Mini, and the Gemini 2.5 / 3.1 family.