For AI agents · MCP

Let your AI make the store screenshots

Built your app with an AI agent? Let it finish the listing too. Add Screenshot Studio to Claude Code, Cursor or any MCP client, and your agent turns your app’s real screenshots into App Store and Google Play images — designed, checked and exported into your project. Free, and it all happens on your computer.

Install

Claude Code — run in your project:

claude mcp add screens -- npx -y fndry39-screens

Then ask: “Make App Store screenshots for this app.” In Claude Code, /mcp__screens__store_screenshots walks the whole job.

Other AI tools

Any MCP client that runs local servers works. Paste the snippet into the file shown and restart the tool.

  • Cursor

    .cursor/mcp.json (or ~/.cursor/mcp.json for every project)

    {
  "mcpServers": {
    "screens": { "command": "npx", "args": ["-y", "fndry39-screens"] }
  }
}

  • VS Code

    .vscode/mcp.json

    {
  "servers": {
    "screens": { "type": "stdio", "command": "npx", "args": ["-y", "fndry39-screens"] }
  }
}

  • Windsurf

    ~/.codeium/windsurf/mcp_config.json

    {
  "mcpServers": {
    "screens": { "command": "npx", "args": ["-y", "fndry39-screens"] }
  }
}

  • Claude Desktop

    Settings → Developer → Edit Config (claude_desktop_config.json)

    {
  "mcpServers": {
    "screens": { "command": "npx", "args": ["-y", "fndry39-screens"] }
  }
}

  • Codex CLI

    ~/.codex/config.toml

    [mcp_servers.screens]
command = "npx"
args = ["-y", "fndry39-screens"]

What your agent gets

  • Ask in plain words

    Your agent already knows what your app does. It takes the screenshots, writes the captions, picks a template and makes every size — you say what you want changed.

  • It looks before it ships

    Every change comes back with a picture of the slides and the studio’s check: a headline running into the phone, a missing translation, a store rule. The agent fixes them before it exports.

  • Files where uploads expect them

    Exact store sizes, 24-bit PNGs, foldered by store, language and screenshot slot, with a README saying which folder goes where.

  • Runs on your computer

    No account, no API key, no upload. It draws in a Chrome on your machine with the page cut off from the network — and it can screenshot your localhost dev server, which no online tool can reach.

Things to ask for

  • “Make App Store screenshots for this app — iPhone and iPad, in our brand blue.”
  • “Now do the Google Play set too, with a feature graphic, in English and French.”
  • “Make a 1200×630 link preview of localhost:3000 with the headline “Ship it [faster]”.”
  • “Translate our store screenshots into German and Japanese, check them, and export.”

The tools

  • screens_list_templates

    The templates, with their typefaces and colours.

  • screens_store_requirements

    Every size, count and file rule the stores publish, and the 30 language codes.

  • screens_create_project

    Designs the set from your screenshots and captions, and returns a picture of every slide and the check.

  • screens_update_project

    Captions, translations, screenshots, slides, devices, template, colours, fonts.

  • screens_preview

    Any device in any language, side by side, to look at before exporting.

  • screens_inspect_project

    Reads a project — including one you made in the studio.

  • screens_export

    The store-ready files, foldered by store, language and slot, with a README.

  • screens_capture_website

    Screenshots a web page in a desktop, tablet or phone viewport — localhost included.

  • screens_make_mockup

    A website in a browser window: a 16:9 slide, a link preview or a social post.

Same studio, same pixels

The server is built from the studio’s own code, so your agent works with the same templates, frames, typefaces and checks you see here. What it makes is an ordinary studio project: open it with Import backup to nudge a headline by hand, then let the agent export again.

  • iPhone, iPad, Android phone and tablet, and the Play feature graphic
  • Captions in 30 languages, with a screenshot per language where the app is translated
  • Browser mockups as slides, link previews and social posts
Open the studio See the templates

Questions

What is an MCP server?

MCP, the Model Context Protocol, is the standard way AI coding tools are given new abilities. Add this server to your tool and your agent can make store screenshots and browser mockups the way it already reads files or runs tests — you ask in plain words, it does the work.

Does it upload my screenshots?

No. The server runs on your computer and draws everything in a Chrome there; the studio page it uses is cut off from the network, so your screenshots and captions go nowhere. Two things do touch the internet: a website you ask it to capture is visited like any page you open, and if your computer has no Chrome, Edge or Chromium, Google’s headless Chrome is downloaded once.

Is it free? Is there an API key?

It is free, with no account, no API key and no limits, like the studio itself. FNDRY39’s terms cover it, and the images you make with it are yours to use, commercially too.

What does it need?

Node.js 20 or newer, which runs npx, and a Chromium-based browser — Chrome, Edge, Chromium or Brave. If none is installed it fetches Google’s headless Chrome, about 100 MB, the first time it is needed.

Are the images the same as the studio’s?

Yes. The server is built from the studio’s own code and draws with the same browser engine, so the same project gives the same pixels. Its projects are ordinary studio backup files: open one in the studio with Import backup to finish it by hand, or hand the agent a backup you made in the studio to translate or re-export.

Where do the files go?

Wherever you or the agent say, inside your project or anywhere else on your computer. Exports are foldered by store, language and screenshot slot, with a README saying which folder goes where. Nothing is overwritten unless the agent is told to.

Will the stores accept what my agent makes?

The server runs the same checks as the studio — exact pixel sizes, PNG or JPEG, no transparency, how many screenshots each store takes, text running into the device — and refuses to export while anything blocks a store upload. Both stores also review what screenshots show, so no tool can promise acceptance.

My agent can’t start it on Windows.

Some Windows tools can’t run npx directly. Use the command cmd with the arguments /c npx -y fndry39-screens instead — for Claude Code: claude mcp add screens -- cmd /c npx -y fndry39-screens.