remark-mcp
Enables AI assistants to browse and read documents from a reMarkable tablet connected via USB, rendering PDF pages as images for the AI to see.
README
remark-mcp
An MCP (Model Context Protocol) server that connects AI assistants to a reMarkable tablet over its local USB HTTP interface.
Once configured, you can ask Claude (or any MCP-capable client) to browse your tablet's documents and read them — the server fetches the PDF from the tablet and renders each page as an image that the AI can see.
Prerequisites
- Node.js 18 or later
- A reMarkable tablet connected via USB cable with the USB web interface enabled
- On the tablet: Settings → Storage → USB web interface — make sure it is turned on
- The tablet must be reachable at
http://10.11.99.1(the default USB address)
Installation
git clone <repo-url>
cd remark-mcp
npm install
Connecting to Claude Code
Add the server to your Claude Code MCP configuration. In your project's .claude/settings.json (or the global ~/.claude/settings.json):
{
"mcpServers": {
"remarkable": {
"command": "/absolute/path/to/node",
"args": ["/absolute/path/to/remark-mcp/index.js"],
"env": {
"RM_MCP_LOG": "info",
"RM_MCP_LOG_FILE": "/absolute/path/to/remark-mcp/rm-mcp.log"
}
}
}
}
Important: The
commandfield must be the full path to thenodeexecutable — Claude Code does not inherit your shell'sPATH, so a barenodewill fail. Find the path withwhich node(macOS/Linux) orwhere node(Windows).
Restart Claude Code after saving. You should see remarkable listed as an available MCP server.
Available Tools
| Tool | Description |
|---|---|
list_root |
List all files and folders in the root of the tablet |
list_folder |
List the contents of a folder by its ID |
download_pdf |
Download a document as a PDF and return each page as an image |
Example prompts
List everything on my reMarkable tablet.
Open the folder called "Work" and show me what's inside.
Show me the document "Meeting Notes" from my tablet.
The download_pdf tool renders every page of the document as a PNG image at 2× viewport scale, so Claude can read handwritten notes and printed text.
Environment Variables
| Variable | Values | Default | Description |
|---|---|---|---|
RM_MCP_LOG |
debug, info, error, silent |
silent |
Log verbosity |
RM_MCP_LOG_FILE |
file path | (stderr) | Write logs to a file instead of stderr |
Note: stdout is reserved for the MCP protocol wire format. All logging goes to stderr or to
RM_MCP_LOG_FILE. Never redirect stdout to a file when running the server.
Running
The server uses the MCP stdio transport. It is normally launched automatically by the MCP client (Claude Code), but you can start it manually to test:
npm start
# or
node index.js
Utility Script
fetch-pdf.js is a standalone script for downloading a single PDF directly to disk — useful for debugging connectivity before using the MCP server:
node fetch-pdf.js <document-id>
# saves <document-id>.pdf in the current directory
Tests
npm test
Tests use Node's built-in test runner (node:test) and stub out all external dependencies (network, PDF renderer, MCP SDK) so they run without a tablet attached.
Windows Notes
The server patches the pdf-to-png-converter path normalisation at runtime so that pdfjs-dist asset URLs use forward slashes on Windows. No manual changes to node_modules are required.
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。