About Pinterest MCP
Pinterest MCP is an open-source MCP server for designers and front-end developers collecting visual references. It searches pins, returns viewable thumbnails and metadata, follows related pins, reads public boards, and downloads original images or videos without a Pinterest login.
Key features
- 01Keyword search with pagination
- 02Viewable pin thumbnails
- 03Pin details and metadata
- 04Related pins and public boards
- 05Image and MP4 downloads
Use cases
- 01Gather UI and visual-design references
- 02Compare layouts and color palettes
- 03Explore pins related to a reference
- 04Download images or motion examples
MCP support
Provides Pinterest search, pin, board, preview, and download tools to MCP clients. The README specifies no login or API key requirement.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @sunkimkk/pinterest-mcpMCP_PINTEREST_DOWNLOAD_DIRFolder for downloaded images (default ~/Downloads/pinterest-mcp)MCP_PINTEREST_PROXYHTTP proxy such as http://127.0.0.1:7890, for networks where pinterest.com is blockedLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
6Saved tool list · updated 1w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| limit | integer | No | How many pins to return (1-100, default 20) |
| media | enum | No | Only return video pins ("video"), only still images ("image"), or both (default)any | video | image |
| query | string | Yes | Search keywords |
| bookmark | string | No | Pass next_bookmark from a previous call to get the next page |
| previews | integer | No | Also attach small thumbnails of the first N pins so you can actually look at them (0-16, default 0). Use this when judging visual style, layout or colour matters. |
Details
From the project READMEPinterest MCP is an open-source MCP Server that gives Claude Code, Codex, Cursor and any other MCP client direct access to Pinterest — search pins, actually look at the images, follow "more like this", read boards, and download originals. It's built for designers and front-end developers who start every task by collecting references.
No Pinterest login. No API key. No headless browser.
❌ Without Pinterest MCP
- You switch to the browser, search, scroll, save images by hand, and drag them back into the chat.
- The AI only gets the images you remembered to paste, without their source, colours or context.
✅ With Pinterest MCP
> Find 20 dark dashboard UIs on Pinterest, look at the first 8,
pick the 3 with the most restrained palette and download them.
The agent searches, looks at the thumbnails, explains its picks, saves full-resolution files to disk — and can go straight on to build the screen from them.
Features
- ⚡ Fast — talks to the same JSON endpoints pinterest.com uses; ~2–3 s for 20 pins, no Chrome to launch.
- 👀 The model can see — optional thumbnails are returned as images, so the agent judges style, layout and colour, not just titles.
- 🧾 Rich metadata — title, description, full-size URL and dimensions (thumbnail-only pins flagged
low_res), dominant colour, source site, board, creator, date. Ads are filtered out. - 🔁 Go deeper — related pins, whole boards, pagination with bookmarks.
- 🛡️ Robust — keeps an anonymous session and retries Pinterest's occasional "fake end of feed".
- 🪶 Token-friendly — hashtag walls collapsed, unusable fields dropped.
- 🎬 Videos too — video pins come with an MP4
video_url; filter to videos only for motion references. - 📥 Downloads — image originals with automatic 736px fallback and MP4 videos, only from
*.pinimg.com.
Quick start
Requires Node.js ≥ 20.18.
Claude Code
claude mcp add pinterest --scope user -- npx -y @sunkimkk/pinterest-mcp
Codex — ~/.codex/config.toml
[mcp_servers.pinterest]
command = "npx"
args = ["-y", "@sunkimkk/pinterest-mcp"]
Claude Desktop — one-click install
Download pinterest-mcp.mcpb from the latest release and double-click it. Claude Desktop shows an install dialog where you can pick a download folder and an optional proxy.
Cursor / Claude Desktop / Windsurf / others — JSON config
{
"mcpServers": {
"pinterest": {
"command": "npx",
"args": ["-y", "@sunkimkk/pinterest-mcp"]
}
}
}
Cursor: ~/.cursor/mcp.json · Claude Desktop: claude_desktop_config.json
From source
git clone https://github.com/sunkimk/pinterest-mcp.git
cd pinterest-mcp && npm install
claude mcp add pinterest --scope user -- node "$PWD/src/server.js"
Tools
| Tool | What it does |
|---|---|
pinterest_search | Keyword search in any language, with pagination and optional thumbnails |
pinterest_get_pin | Full details of one pin (pin URL, pin.it short link or numeric id) |
pinterest_related_pins | "More like this" for a pin |
pinterest_board_pins | Pins saved on a public board |
pinterest_preview_images | Turn image URLs into thumbnails the model can see |
pinterest_download | Save full-resolution images and MP4 videos to disk |
search, related_pins and board_pins accept previews: N to attach thumbnails of the first N results, and media: "video" to return only video pins (each with an MP4 video_url) — handy for motion and interaction references.
Example prompts
- "Search Pinterest for glassmorphism login screens and show me 8 thumbnails."
- "What do these have in common — layout, palette, typography?"
- "Find pins similar to https://www.pinterest.com/pin/1128714725389331478/"
- "List everything on https://www.pinterest.com/awsmd_com/app-design/"
- "Find 10 profile page animation videos and download them."
- "Download #2, #5 and #7 to ~/Desktop/refs/login."
Tip: English keywords return far more results. Add words like
ui,app,dribbble,posterormoodboardto steer the style.
Configuration
| Variable | Default | Description |
|---|---|---|
MCP_PINTEREST_DOWNLOAD_DIR | ~/Downloads/pinterest-mcp | Download root |
MCP_PINTEREST_PROXY | system HTTPS_PROXY | HTTP proxy, e.g. http://127.0.0.1:7890 (SOCKS not supported) |
MCP_PINTEREST_LOCALE | en-US | Request language |
MCP_PINTEREST_TIMEOUT_MS | 20000 | Per-request timeout |
Example with Claude Code:
claude mcp add pinterest --scope user -e MCP_PINTEREST_PROXY=http://127.0.0.1:7890 -- npx -y @sunkimkk/pinterest-mcp
FAQ
Why can't I get details for a search result?
Logged-out search results carry anonymous ids that Pinterest won't resolve. Everything they have is already in the search result; get_pin and related_pins need a real pin link (pinterest.com/pin/<number>/).
Network errors?
If pinterest.com is blocked where you are, set MCP_PINTEREST_PROXY.
Will it break? It uses Pinterest's web endpoints rather than the official API, so a Pinterest redesign may require an update. Watch the repo or check the changelog.
Development
npm install
npm test # spins up the server and calls every tool against live pinterest.com
See CHANGELOG.md for release history. Issues and PRs are welcome.
Disclaimer
Not affiliated with, endorsed by, or sponsored by Pinterest. Use it for personal inspiration at a reasonable request rate, follow Pinterest's Terms of Service, and respect creators' copyright — downloaded images are for reference only.
Acknowledgements
Thanks to terryso/mcp-pinterest for first exploring Pinterest over MCP.
Star History
License
MIT © sunkimk
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.