screensdesign

ScreensDesign MCP

Your AI can use the library now

Connect once and your agent searches 2,715+ top-grossing apps while it works: real screens, full UX flows, revenue signals, and App Store creatives. The library stops being a tab you visit and becomes something your tools simply know.

Hosted server: api.screensdesign.com/v1/mcp · 18 tools · ~2 minutes to connect

Ask in plain language, get the actual screens

“How do the top sleep apps run their paywalls? Show the actual screens.”

Real paywall screens with revenue next to each app

“Pull the full onboarding flow of the highest-grossing habit tracker.”

Screen-by-screen flow, in order, with deep links

“Which finance apps reframe an annual price as price per day?”

The exact plan sheets that use the pattern

“Find discount screens that use a spin-the-wheel mechanic.”

Store and in-app screens matched by meaning, not keywords

“What do quiz-style onboardings convert into? Compare three examples.”

Flows plus the revenue signals behind them

“Generate four app screens in the style of the top meditation apps.”

New screens grounded in what already converts

Works where you already work

ClaudeDesktop & Claude Code
CursorAgent & chat
ChatGPTConnectors
CodexCLI & IDE
WindsurfCascade
Any MCP clientStandard streamable HTTP

Eighteen tools, four jobs

Search

Apps, recorded screens, complete flows, App Store screenshots, and the broader app market — semantic search that understands “hard paywall” and “streak saver”, not just keywords.

Evidence

App detail with revenue and install signals, screen detail, similar-screen lookup, and review analytics across countries — uncover recurring pain points, feature requests, sentiment, and the patterns users mention most.

Collections

Your saved research: list collections and pull the screens inside them straight into the conversation.

Generation

Generate app screens grounded in the library’s evidence — your agent studies what converts before it draws.

Questions, answered

What is the ScreensDesign MCP server?

It is a hosted Model Context Protocol server at api.screensdesign.com/v1/mcp. MCP is the open standard AI assistants use to reach external tools, so once connected, your AI can search the ScreensDesign library — 2,715+ top-grossing iOS apps, recorded screens, full UX flows, revenue signals, and App Store creatives — while it answers you.

Which AI tools work with it?

Claude (Desktop and Claude Code), Cursor, ChatGPT, Codex, Windsurf, and any client that speaks standard MCP over streamable HTTP. Setup takes about two minutes per client.

Is it included with a Pro subscription?

Yes. MCP access is part of ScreensDesign Pro — there is no separate plan. You authorize your client with your ScreensDesign account or a scoped API key.

What can my agent actually do with it?

Eighteen tools across search, evidence, collections, and generation: find apps and screens by meaning, open full flows, check revenue and install signals, read App Store reviews, browse your saved collections, and generate new screens grounded in the library.

How is access secured and rate limited?

Connections authorize through OAuth or scoped API keys tied to your organization. Fair-use budgets bound request rates and data volume; normal research usage never touches them.

How do I connect?

Open the setup page, pick your client, and follow the two-minute walkthrough — it generates the exact config for Claude, Cursor, ChatGPT, or Codex and verifies the connection live.

Two minutes from now, your agent knows the library