ui-screenshot-mcp
Searches Dribbble shots as a design-inspiration source, returning shot results that can be downloaded into the local library. Dribbble search requires a Playwright Chromium browser.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@ui-screenshot-mcp find Duolingo onboarding screens and save them to my library"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
ui-screenshot-mcp
An MCP server and CLI for UI research. Search real app and website screens from public design libraries, download them into a local library that is organized by category, search that library later, and screenshot any web page with a real browser.
Works with Claude Code, Claude Desktop, Gemini CLI, Codex and any other MCP client.
Features
Search design libraries: Refero (iOS and web), Mobbin (iOS, Android, web), ScreensDesign and Dribbble.
Build a local library: download a whole app or a search into
<library>/<category>/<app>/. Downloads resume, and screens locked behind a paywall are skipped.Search offline: each image has a
.jsonfile with its source, app and tags, andlibrary_searchmatches against it.Capture web pages: full page, single element or viewport, with waits for animations and a way to hide cookie banners.
Related MCP server: Mobbin Agent
Install
Claude Code plugin (recommended)
/plugin marketplace add Angioz/ui-screenshot-mcp
/plugin install ui-screenshot-mcp@angiozThis installs the MCP server and the ui-screenshot-mcp skill, which teaches Claude how to use the tools. Claude Code asks where to keep the library; leave it empty for ~/.ui-library.
Any MCP client
{
"mcpServers": {
"ui-screenshot": {
"command": "npx",
"args": ["-y", "ui-screenshot-mcp"],
"env": { "UI_LIBRARY_DIR": "/path/to/your/ui-library" }
}
}
}Or with the Claude Code CLI:
claude mcp add --scope user ui-screenshot -- npx -y ui-screenshot-mcpTo use the skill in other harnesses, copy skills/ui-screenshot-mcp/ into their skills folder.
Browser
Web capture and Dribbble search need a Playwright browser. Install it once:
npx playwright install chromiumConfiguration
Variable | Default | Purpose |
|
| Where downloaded screens are saved |
|
| Mobbin login cookies (see below) |
Mobbin login
Mobbin only serves screens to logged-in users. To use it:
Log in at mobbin.com in your browser.
Export the site's cookies in Netscape
cookies.txtformat, for example with the "Get cookies.txt LOCALLY" extension.Save the file as
mobbin-cookies.txtin your library folder, or pointMOBBIN_COOKIESat it.
The file is a login: keep it private and never commit it. Free plans only open part of each app, and locked screens are skipped. When the login expires, export the cookies again.
Sources
Source id | Search returns | Account |
| screens | none |
| apps | none |
| apps | Mobbin login |
| shots | none |
Use the appId from an app result with get_ui_screens or download_ui.
Library layout
~/.ui-library/
├── 01-onboarding/
│ └── duolingo/
│ ├── 6d1f….webp
│ └── 6d1f….webp.json
├── 06-paywalls/
└── 99-other/The category comes from the search query when it names one, otherwise from the screen's first matching tag. The categories are:
01-onboarding, 02-authentication, 03-main-feed, 04-profile, 05-dashboard, 06-paywalls, 07-navigation, 08-search-discover, 09-saved-collections, 10-settings, 11-ai-chat, 12-quiz-gamification, 13-empty-states, 14-notifications, 15-progress, 16-loading, 17-dark-mode and 18-selection. Anything that matches none of them goes to 99-other.
MCP tools
Tool | What it does |
| List sources; with |
| Search a source ( |
| List one app's screens with tags ( |
| Save an app ( |
| Search the local library by words |
| Library folder, image count and size per folder |
| Screenshot a URL or an open session ( |
| Keep a browser session open for several captures |
| List captured screenshots |
CLI
The same features without an MCP client:
npx ui-screenshot-mcp # MCP server on stdio
npm i -g ui-screenshot-mcp # installs the `ui-screenshot` CLI
ui-screenshot sources --check
ui-screenshot search --source refero-ios --query paywall
ui-screenshot screens --source screensdesign --app 862
ui-screenshot sync --source refero-ios --query onboarding --limit 50
ui-screenshot library --query "duolingo paywall"
ui-screenshot capture --url https://example.com --full-page --output page.png
ui-screenshot --helpDevelopment
git clone https://github.com/Angioz/ui-screenshot-mcp
cd ui-screenshot-mcp
npm install
npx playwright install chromium
npm testDisclaimer
This project is not affiliated with Refero, Mobbin, ScreensDesign or Dribbble. It uses their public web endpoints, which can change without notice. Respect each site's terms, keep downloads for personal reference and research, and don't redistribute the images.
License
MIT © Angioz
Available Tools
10 toolscapture_screenshotB
Capture a screenshot of a webpage or current page. Returns path to saved screenshot.
Inputs: url, width, format, height, +13 more.
close_browserC
Close a browser session
Inputs: sessionId.
download_uiA
Download screens into the local UI library (UI_LIBRARY_DIR, default ~/.ui-library): every screen of an appId, or the results of a query. Already-downloaded and paywalled screens are skipped.
Inputs: source (required), appId, limit, query.
get_ui_screensA
List the screens of one app/site (appId from search_ui results), with image URLs and tags
Inputs: appId (required), source (required), page.
library_searchB
Search the local UI library by words matched against folders, file names and saved metadata (app, tags, source)
Inputs: query (required), limit.
library_statsB
Count the images in the local UI library per source/app folder
Inputs: no named parameters.
list_screenshotsC
List all screenshots in the output directory
Inputs: directory.
open_browserA
Open a browser and navigate to a URL. Returns page ID for subsequent screenshots.
Inputs: url, width, height, browser, +8 more.
search_uiA
Search a UI library source. Returns screens (refero-ios, refero-web, dribbble) or apps (screensdesign, mobbin-*); pass an app's appId to get_ui_screens or download_ui.
Inputs: query (required), source (required), page.
ui_sourcesA
List the UI library sources; with check=true, run a live search on each to report which ones work right now
Inputs: check.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 10 tools
Most tools have distinct purposes: external source exploration (ui_sources, search_ui, get_ui_screens, download_ui), local library management (library_search, library_stats), and browser capture (open_browser, capture_screenshot, close_browser, list_screenshots). The two search tools (search_ui vs library_search) and two image-saving tools (download_ui vs capture_screenshot) could be momentarily confused, but descriptions clarify remote vs local and download vs capture.
The set mixes verb_noun names (capture_screenshot, open_browser, list_screenshots) with noun_phrases (ui_sources, library_stats) and uses inconsistent prefixes like 'ui_' and 'library_' for related concepts. It remains readable, but the convention is not predictable across all tools.
Ten tools is well-scoped for a server that covers source discovery, search, screenshot capture, download, and local library management. Each tool has a clear role in the workflow without unnecessary duplication.
The surface covers the main lifecycle: discovering sources, searching remote and local libraries, downloading screens, opening a browser, capturing screenshots, and listing outputs. Minor gaps exist, such as no explicit tool to delete or update local library entries, but core workflows are covered.
Related MCP Connectors
Web MCP: scrape/crawl sites, web search, brand assets, app stores, YouTube, Reddit, Hacker News.
Web scraping, browser automation, screenshots and AI data extraction for AI agents via MCP.
Clean PNG/JPEG screenshots via REST or MCP, with goal-driven multi-step navigation.
One MCP server for all web data. Scrape, search, crawl, and AI research in one install.
Related MCP Servers
- AlicenseAqualityCmaintenanceMCP server that scrapes Terminal Trove to search CLI/TUI tools and retrieve design references like screenshots and install commands for agentic UI research.12MIT
- FlicenseAqualityDmaintenanceMCP server that gives AI agents browser-based access to Mobbin for design research, screen discovery, and screenshot collection through automated browser control.103-
- AlicenseNot gradedqualityCmaintenanceAn MCP server that works with Mobbin — 600k+ real app screens and flows from 1,100+ apps. Search screens, browse flows, grab palettes, build better UI without leaving Claude.6ISC
- AlicenseNot gradedqualityAmaintenanceScreenshot any URL from the command line with metadata sidecars, accessibility audits, visual regression, performance analysis, and MCP integration for AI-assisted development.34 npm1MIT