mdlog-mcp
An MCP server that provides read-only tools to search, list, and read date-organized markdown logs, enabling natural-language queries over past notes and digests.
README
mdlog-mcp
MCP (Model Context Protocol) server for searching date-organized markdown logs — daily digests, journals, meeting notes, or any markdown corpus where dates live in filenames or directory names.
Point it at a directory and ask your AI agent things like "What did my digest say about Next.js caching back in May?" — the agent finds it via full-text search instead of you grepping by hand.
Recognized layouts
Any markdown file whose date appears in the filename or a parent directory name becomes an entry:
logs/
├── 2026-04-24-ai-web-dev-digest.md # date in filename
├── 2026-08-10/
│ ├── ai-web-dev-digest.md # date in directory name
│ └── infra-security-data-digest.md
└── weekly/
└── 2026-08-10-weekly-review.md # works in subdirectories too
- date:
YYYY-MM-DDfrom the filename prefix or the nearest dated directory (must be a real calendar date) - slug: the rest of the filename (e.g.
ai-web-dev-digest) — acts as the entry type - category: the subdirectory (e.g.
weekly), with date directories stripped - Undated markdown files are ignored and cannot be read through the tools
- Symlinks inside the directory are never followed
Tools
All tools are read-only.
| Tool | Purpose |
|---|---|
mdlog_overview |
Corpus summary: entry count, date range, types, latest entries. Call first. |
mdlog_list_entries |
List entries newest-first, filtered by date range / slug / category, paginated. |
mdlog_search |
Full-text AND search with line-numbered snippets and context lines. |
mdlog_get_entry |
Read one entry's full content, chunked for long files. |
Every tool supports response_format: "markdown" | "json" (where applicable) and returns structuredContent alongside text.
Setup
Requires Node.js >= 20.
Claude Code
claude mcp add --scope user mdlog -- npx -y @yamk12nfu/mdlog-mcp --dir /path/to/your/logs
Or from a local checkout:
npm install && npm run build
claude mcp add --scope user mdlog -- node /path/to/mdlog-mcp/dist/index.js --dir /path/to/your/logs
Claude Desktop / other MCP clients
{
"mcpServers": {
"mdlog": {
"command": "npx",
"args": ["-y", "@yamk12nfu/mdlog-mcp", "--dir", "/path/to/your/logs"]
}
}
}
The target directory can also be set via the MDLOG_DIR environment variable instead of --dir.
Example prompts
Once connected, ask your agent:
- "5月ごろのダイジェストで Next.js のキャッシュの話なかったっけ?"
- "What security incidents were covered in my digests between June and July?"
- "Summarize the last three weekly reviews."
Development
npm install
npm run build # compile to dist/
npm run test # run the test suite (node:test via tsx)
npm run dev # run from src/ with tsx
node dist/index.js --dir ./sample-logs
Debug interactively with the MCP Inspector:
npx @modelcontextprotocol/inspector node dist/index.js --dir /path/to/your/logs
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。