raindrop-mcp
Enables management of Raindrop.io bookmarks, collections, tags, and highlights via MCP tools, with support for search, bulk editing, and library auditing.
README
raindrop-mcp
A Raindrop.io MCP server in Python (FastMCP), serving streamable HTTP natively — no stdio↔HTTP bridge needed. Also runs as a plain stdio server for clients that spawn a local process.
Built as a drop-in replacement for
@adeze/raindrop-mcp: the 17 tools keep
the same names, parameters, and required fields, so existing prompts and client
configs keep working. (Two tools differ internally: get_suggestions /
suggest_tags use Raindrop's native suggest endpoint rather than
MCP-sampling-based AI refinement.)
Tools
| Tool | What it does |
|---|---|
collection_list |
All collections, flat |
get_collection_tree |
Collections as a hierarchy with breadcrumb paths |
collection_manage |
Create / update / delete a collection |
cleanup_collections |
Remove empty collections (dry-run unless confirm: true) |
bookmark_search |
Full-text search with tag / domain / type / date / flag filters |
bookmark_manage |
Create / update / delete a bookmark |
get_raindrop |
One bookmark by ID (with highlights) |
list_raindrops |
Bookmarks in a collection, paginated |
bulk_edit_raindrops |
Bulk update / move / remove |
library_audit |
Count (and list) broken links and duplicates |
empty_trash |
Permanently empty Trash (dry-run unless confirm: true) |
remove_duplicates |
Delete same-URL duplicates, keeping the oldest (dry-run by default) |
tag_manage |
Rename / merge / delete tags |
highlight_manage |
Create / update / delete highlights |
get_suggestions |
Suggested tags + collections for a URL or bookmark |
suggest_tags |
Suggested tags for a URL |
diagnostics |
Server/runtime metadata |
Collection IDs: 0 = all, -1 = Unsorted, -99 = Trash. Destructive sweeps
default to a dry run.
Setup
Requires Python ≥ 3.10 and uv.
uv sync
cp config/.env.example config/.env # then paste your token
Get a token at raindrop.io → Settings → Integrations → For Developers →
create an app → Test token. Set it as RAINDROP_ACCESS_TOKEN in
config/.env (or in the process environment, which takes precedence).
Running
Streamable HTTP (network clients — Claude Desktop remote MCP, gateways):
MCP_TRANSPORT=http MCP_PORT=8591 uv run start
# serves http://0.0.0.0:8591/mcp
stdio (clients that spawn the server locally):
uv run start
systemd user unit (example)
[Unit]
Description=Raindrop.io MCP (native FastMCP HTTP)
After=network.target
[Service]
Type=simple
WorkingDirectory=%h/raindrop-mcp
Environment=MCP_TRANSPORT=http
Environment=MCP_HOST=0.0.0.0
Environment=MCP_PORT=8591
ExecStart=%h/raindrop-mcp/.venv/bin/start
MemoryMax=512M
MemorySwapMax=0
Restart=always
RestartSec=5
[Install]
WantedBy=default.target
The server binds 0.0.0.0 in the example — front it with your LAN/VPN
(Tailscale works well); the Raindrop token authorizes every request it serves.
Notes
- Scaffolded with mcp-template.
- Rate limiting: honours
Retry-Afteronce on HTTP 429 (Raindrop allows 120 req/min). skipCacheparameters are accepted for compatibility but are no-ops — this server doesn't cache.
License
MIT
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。