Savee MCP Server
Enables conversational access to your saved visual inspiration on Savee, including listing saves, boards, feeds, searching the public library, and viewing images, with read-only tools.
README
Savee MCP Server
Your saved visual inspiration, in the conversation.
Savee is a curated visual-inspiration platform for designers and creative professionals. Members save images and video from anywhere on the web into boards, and discover work through color, tags, and visual similarity.
Savee runs a hosted remote MCP server at https://mcp.savee.com/mcp. There
is nothing to clone, install, or run locally — this repository holds the
plugin manifest, the registry manifest, and setup docs. The server itself is
closed-source and operated by Savee.
Setup
Add the server to your MCP client:
{
"mcpServers": {
"savee": {
"type": "streamableHttp",
"url": "https://mcp.savee.com/mcp"
}
}
}
The transport is Streamable HTTP. Some clients — Cline among them — fall back to
the legacy SSE transport when type is omitted, which fails to connect; setting
"type": "streamableHttp" explicitly avoids that. Clients that only accept a URL
(Claude, VS Code) negotiate the transport themselves and need nothing extra.
Claude Code
claude mcp add --transport http --scope user savee https://mcp.savee.com/mcp
Then run /mcp, select savee, and choose Authenticate.
Cursor
Settings → MCP → Add new MCP server, pick the HTTP type, and paste
https://mcp.savee.com/mcp. Or drop the snippet above into ~/.cursor/mcp.json.
Claude
Settings → Connectors → Add custom connector, and paste
https://mcp.savee.com/mcp.
Everything else
Per-client instructions live at docs.savee.com/mcp/install.
Authorization
Authentication is OAuth 2.1 with PKCE — Client ID Metadata Documents preferred, Dynamic Client Registration supported. On first connection your browser opens savee.com, you sign in, and a consent screen lists exactly what you're granting. There is no API key to copy anywhere in the flow.
You can disconnect at any time from Settings → Developers on savee.com, which takes effect immediately.
Tools
All seven tools are read-only and annotated readOnlyHint. No write tools
exist, so nothing here can save, edit, move, or delete anything in your account.
| Tool | What it does |
|---|---|
get_current_user |
The signed-in Savee profile |
list_saves |
Your own saves, newest first |
list_feed |
Your home feed |
list_boards |
The boards you own or belong to |
list_board_saves |
The saves on one board |
search |
Savee's public library — everything on the platform, not just your saves |
view_saves |
The actual images for specific saves, so the assistant can see them |
Every save comes back with its source URL, dimensions, and dominant colors, so
an assistant can reason about what's in a collection rather than just list
links. view_saves goes further and returns the images themselves, for
questions about composition, layout, type, or how a set of images sit together.
Full reference: docs.savee.com/mcp.
What it can see
Private saves and private boards are included — the connector sees what you see when signed in. The Savee setting "Hide private board saves from profile feed" applies here too. Scopes are granted individually on the consent screen, and a tool whose scope you declined returns an error rather than data.
Requirements
A Savee account with an active subscription. Check your plan.
Links
- Documentation — https://docs.savee.com/mcp
- Privacy policy — https://savee.com/privacy/
- Support — hey@savee.com
License
The contents of this repository are MIT licensed. Use of the hosted service is governed by Savee's terms.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。