Automate store screenshots with AI agents (WebMCP)

WebMCP lets a web page expose tools to AI agents running in the browser. Instead of clicking through the UI, an agent calls typed functions — manage_screens, set_copy, export — and the editor updates live while you watch.

What an agent can do

• Create a project, add platforms (iPhone, iPad, Android, Mac, TV…) and upload simulator screenshots as data URLs.
• Look at the captures (get_images with mode: "original"), write headlines, and translate them into every project language in one set_copy call.
• Style backgrounds, device frames (2D chrome or 3D iPhone/Galaxy), typography and per-language layouts.
• Create A/B variants, copy designs across platforms, and compare them.
• Check its own work: get_images renders the real composition; get_screen reports overlaps and overflowing text.
• Export full-resolution PNGs per variant, platform and language — or build a ZIP for you.

A typical run

1. get_app_state                      → projects, variants, platforms, selection, hints
2. manage_project  { action: "create", name: "Veena", platformIds: ["app-store-iphone", "google-play-phone"] }
3. manage_screens  { action: "upload", screenshots: [{ filename: "01-dictate.png", dataUrl: "data:image/png;base64,…" }] }
4. get_images      { mode: "original", scope: "platform" }
5. set_copy        { entries: [{ screenIndex: 0, field: "headline", language: "en", text: "Just speak. Veena types." }, …] }
6. set_background  { scope: "variant", gradientPreset: "Morning Mist" }
7. set_text_style  { scope: "variant", headline: { font: "Plus Jakarta Sans", weight: "800", color: "#1c1c28" } }
8. get_images      { scope: "platform" }   → review, fix warnings from get_screen
9. export          { allLanguages: true, allPlatforms: true, delivery: "return" }

Using it from a coding agent

Browser-automation agents (for example agent-browser or a WebMCP-enabled Chrome) can open storedeck.byanr.com and call the tools directly. For agents that can only evaluate JavaScript, the same tools are available as window.storedeck.callTool(name, input).

Add the snippet from the agents page to your repository's AGENTS.md so every coding agent knows the workflow: capture simulator screenshots, open Storedeck, design, ask you for feedback, export, and save the files next to your code.