The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the ScreensDesign listing page.
Give your AI agent real app screens, complete flows, and market context to work from.
The ScreensDesign MCP is a hosted server for researching mobile apps, comparing onboarding and paywalls, exploring App Store reviews, and finding visual references. Connect it to your MCP client to bring real screens, recorded flows, and market context into your research.
This public repository contains setup documentation, registry metadata, and the optional screensdesign-data companion skill. The hosted server implementation and ScreensDesign data are operated separately.
Website · Connect the MCP · Browse on skills.sh · Releases
Add this hosted endpoint to your MCP client using Streamable HTTP and sign in with your ScreensDesign account:
MCP access is included with ScreensDesign Pro. See the setup page for supported clients and access details.
You can also open /mcp inside Claude Code to check the connection and authenticate.
Add this server to your Cursor MCP configuration:
Complete the browser sign-in when prompted.
Authentication uses browser OAuth. If the tools do not appear after setup, refresh the MCP connection or start a new conversation. See connection troubleshooting for more detail.
The server provides tools for researching apps, screens, flows, reviews, and saved collections:
| Research task | What you get |
|---|---|
| Discover apps and competitors | Apps matched by product, category, positioning, or recorded UI, with revenue and download estimates where available. |
| Study onboarding and paywalls | Recorded screens, chronological replays, and stored flows that show how a journey unfolds. |
| Find UI references | Relevant screens across apps, focused searches within an app, and visual matches to a reference image. |
| Explore the broader app market | Competitor discovery, public app listings, and App Store reviews for researching recurring complaints and positioning. |
| Research App Store creatives | Marketing screenshots searched separately from recorded in-app screens. |
| Continue saved research | Your saved app collections and developer portfolios brought into the conversation. |
The optional screensdesign-data skill guides agents to check replay context before making claims about screen order and link findings to the supplied app, screen, or flow sources. It distinguishes observed UI from inference and treats revenue and downloads as performance signals, not proof of conversion.
With the MCP connected, ask your agent:
Compare onboarding in three high-revenue habit trackers. Show the recorded screens in order and where each paywall appears.
Find apps similar to my sleep app idea. Compare their positioning and summarize recurring complaints in their public App Store reviews.
Find paywalls that show an annual subscription as a price per day. Link to the actual screens.
Find app screens visually similar to this screenshot. Explain which layout and navigation patterns are shared.
Compare the App Store screenshot messaging of three meditation apps. Which benefits does each lead with?
Open my saved app collection and compare the onboarding patterns across its apps.
The MCP works without installing a skill. For agents that support skills, screensdesign-data adds guidance on choosing research tools, checking evidence, and citing sources.
Run this from your project root:
Then connect the ScreensDesign MCP. The skill provides research instructions; the MCP provides access to the data through your ScreensDesign account.
Install for specific agents without prompts:
Install across your projects:
Install the exact v1.0.8 release:
Update an installed copy:
server.json describes the hosted ScreensDesign MCP server for the official MCP Registry under com.screensdesign/screensdesign. It includes the public repository, remote endpoint, display name, icon, and setup link. Registry metadata version 1.0.1 is maintained separately from the data-skill release.
Changes to this file must be published to the official MCP Registry with mcp-publisher publish after authenticating for the com.screensdesign namespace. Pushing a Git branch does not update the registry. Verify the repository field and version in the latest registry record after publishing.
The endpoint uses browser OAuth; complete sign-in in your MCP client. This repository contains the companion skill and registry metadata. The hosted server implementation and ScreensDesign data are operated separately.
The skill stays compact and loads focused guides for app research, market and review research, screen and flow research, app intelligence, and saved collections as needed.
The tool reference and response field guide document the research interface. Live MCP schemas take precedence when a tool changes.
The current data-skill release is 1.0.8. When connected, the agent checks its installed version once per conversation through get_screensdesign_skill to learn whether it is current or needs an update.
Each release uses a semantic Git tag matching the version declared in screensdesign-data/SKILL.md. For the current release:
The release manifest records immutable content and ZIP hashes. Pushing the matching tag validates the package and creates a GitHub Release with the ZIP and manifest. The hosted MCP vendors that exact package for authenticated clients to read or download through MCP resources.
MIT © 2026 ScreensDesign. The ScreensDesign name and logo are brand assets; this license does not grant trademark rights.
Built by ScreensDesign — real app screens, recorded flows, and market intelligence for your next product decision.