Let AI agents make your store screenshots

START HERE. Returns the open project (languages, variants → platform decks → screens with ids, image languages and headline/subheadline copy), all saved projects, the current selection and view, undo history and next-step hints. Image bytes are not included (use get_images). Call again after structural changes.

Full design of one screen (default: selected): background, device, text style, copy per language, elements, popouts, image info, output size, plus computed layout (device and text rectangles in output px) and warnings such as text overlapping the device or running off-canvas.

Store screenshot specs: every platform id with its output sizes, orientation, screenshot count limits and requirements. Stores: app-store, google-play, microsoft-store, steam, amazon-appstore, tv-stores, web.

The localization table: every headline, subheadline and text-element string per language, with missing translations listed. Filter by variant/platform. Use before translating, then write with set_copy.

Images for visual QA. mode "rendered" (default) returns the styled composition exactly as exported, downscaled to maxDimension (default 900 px, JPEG). mode "original" returns uploaded source screenshots (use to read the app UI before writing copy). Waits for fonts, images and 3D models. Target with screenId/screenIds/screenIndex or scope.

Moves the editor focus so the user sees what you work on: variantId (A/B variant), deckId (platform), screenId or screenIndex, and language (preview language). Switching variant keeps the same platform and screen position when possible.

Controls the editor UI: view (design = single screen canvas, board = all screens × languages grid, strings = localization table, compare = A/B variants side by side), inspectorTab, selected element/popout, which variants to compare, and the store safe-area overlay.

Projects are the top-level container (one per app). Actions: create {name, appName?, platformIds?, languages?} — new project with an "Original" variant; open {projectId}; rename {name, appName?}; duplicate {projectId?, name?}; delete {projectId, confirm:true} (permanent); export_file {projectId?} downloads a .storedeck.json backup for the user; import_file {bundle} imports a backup (object or JSON string).

Variants are alternative screenshot sets for A/B tests (App Store Product Page Optimization, Google Play store listing experiments). Actions: create {name?, cloneFromVariantId?, blank?, platformIds?} — clones the selected variant by default so you only change what the test is about; update {variantId, name?, status?, hypothesis?} (status draft|control|testing|winner|archived — one control/winner at a time); delete {variantId}; move {variantId, toIndex}.

Each variant holds one deck per store platform (iPhone, iPad, Android phone, Mac, Apple TV, Android TV, Windows, Steam, Fire TV, …; ids via get_platform_catalog). Actions: add {platformId, sizeId?, variantId?, cloneFromDeckId?, allVariants?} — cloneFromDeckId copies screens, images and copy and rescales the design to the new size (the fastest way to go from iPhone to Android/iPad); remove {deckId}; set_size {deckId?, sizeId? | width+height, orientation?} — design values rescale automatically.

Screens are the individual store images of a platform deck. Actions: upload {screenshots:[{filename, dataUrl, language?}], deckId?, screenId?} — PNG/JPEG data URLs, max 10 per call; language suffixes in filenames (home_de.png, home-pt-br.png) are detected, and files whose base name matches an existing screen become its localized image; new screens copy the selected screen's style. add_blank {deckId?, name?}; delete {screenId}; duplicate {screenId}; move {screenId, toIndex} (store order); rename {screenId, name}; remove_image {screenId, language}.

Project languages (lower-case BCP-47 like "de", "pt-br", "zh-tw", "te"). Actions: add {languages:[...]}, remove {language} (copy is kept and returns if re-added), set_default {language} (fallback for missing copy/images), switch {language} (preview language — same as select).

Writes store copy (the words, not styling) in bulk. Each entry targets a screen (screenId, or screenIndex + optional deckId), a field (headline, subheadline, or "element:<elementId>" for text badges), a language and text. Languages are added to the project automatically. Copywriting: headline 2–5 words, benefit-first; subheadline 4–10 words that expand it; every screen tells a different part of the story and screen 1 carries the core value proposition. Use \n for manual line breaks. Prefer one call with all entries.

Copies headline/subheadline copy from one platform deck to another by screen position (e.g. reuse iPhone copy for Android). Optional languages filter.

Background of the targeted screens: type gradient|solid|image; gradient {kind linear|radial, angle (CSS degrees, 0 = up), stops [{color, position 0-100}]}; gradientPreset (Indigo Rush, Midnight Abyss, Obsidian Plum, Carbon Slate, Steel Blue, Neon Horizon, Electric Surge, Synthwave Dusk, Northern Lights, Deep Forest, Emerald Canopy, Ocean Pulse, Desert Dusk, Ember Glow, Mocha Silk, Golden Hour, Pacific Sunset, Volcanic Dawn, Deep Ocean, Reef Lagoon, Gold Noir, Velvet Noir, Morning Mist, Sage Whisper, Royal Navy, Paper White); solid; imageDataUrl (sets type image); imageFit cover|contain|stretch; imageBlur; overlayColor/overlayOpacity (tint for contrast); noise/noiseIntensity.

How the screenshot is presented. preset (centered: Centered; bleed-bottom: Bleed bottom; bleed-top: Bleed top; float-center: Float center; tilt-left: Tilt left; tilt-right: Tilt right; perspective: Perspective; float-bottom: Float bottom; side-right: Text left, device right; side-left: Device left, text right) — explicit fields override it. scale (% of canvas width), x/y (50 = centered; <0 or >100 bleeds off-canvas), rotation, perspective, cornerRadius; chrome {style: none|phone|tablet|laptop|monitor|browser|window|tv|watch, color} draws a 2D device frame (phone for mobile, laptop/window/browser for desktop, tv for TV); use3D with model3D iphone|samsung, modelColor and rotation3D {x,y,z}; shadow {...}; border {...} (outline stroke).

Typography and layout of the headline/subheadline (content is set with set_copy). headline/subheadline {enabled, font, size (output px), weight, color, opacity, italic, underline, strikethrough, uppercase, letterSpacing}; position top|bottom; align left|center|right (mirrored automatically for RTL languages); offsetY (% from edge); x (block center %) and maxWidth (% of width) for side-by-side layouts; lineHeight (%); gap (% of headline size). With language set, headline/subheadline size, offsetY and lineHeight are stored as an override for that language only (for long German or short CJK copy); resetLanguageOverride removes it.

Floating overlays on a screen (badges, awards, emoji, icons, logos — not the headline). Actions: add {kind: text|emoji|icon|image, text+language (text), emoji, icon (Lucide name), imageDataUrl (image), properties?}; update {elementId, properties?, text?, language?}; delete {elementId}; reorder {elementId, toIndex} (later = on top). Layers: behind-device, above-device, above-text.

Popouts crop a region of the screen's source screenshot and show it enlarged (callouts for key UI). Actions: add {properties?}, update {popoutId, properties}, delete {popoutId}. Crop values are % of the source image; placement is % of the canvas.

Copies design from a source screen to targets — across screens, platforms or variants. Copy (words) and screenshots stay with each target; pixel values rescale when sizes differ. parts selects background/device/text/elements (all by default). Targets: targetScreenIds, or targetScope platform (source deck) | variant (every deck in source variant) | project (everything).

Renders full-resolution store images. delivery "return" (default) returns data URLs to save as files (max 24 per call — narrow with screenIds/deckIds/languages). delivery "download" builds a ZIP in the browser for the user (variant/platform-WxH/lang/NN-name.png). Defaults: selected platform deck, current language, PNG. JPEG is flattened (stores reject alpha).

Undo or redo document changes (yours and the user's). steps defaults to 1.