upnote-lens-mcp
A hybrid MCP server that reads UpNote notes from the local SQLite database (returns actual text) and writes/opens notes via the upnote:// URL scheme, enabling search, retrieval, creation, and navigation of notes.
README
upnote-lens-mcp
<!-- mcp-name: io.github.AwesomeHye/upnote-lens-mcp -->
A hybrid MCP server that lets your AI assistant work with your UpNote notes:
- 🔍 Find & search notes by keyword across titles and bodies.
- 📖 Read & summarize — it returns the actual note text, so the AI can summarize, analyze, or answer questions about what you've written.
- ✍️ Create new notes from the chat.
Under the hood: reads come from the local UpNote SQLite database (read-only),
so real content comes back as text; writes go through the upnote:// URL scheme
(x-callback-url) and never touch the database.
The write side (URL scheme) is based on chadthornton/upnote-mcp (MIT).
Usage
Once it's registered, just talk to your AI assistant in plain language — it picks the right tool for you. No commands or syntax to memorize. For example:
- "Summarize my UpNote notes about the Q3 roadmap."
- "Find notes where I mentioned Postgres tuning and pull out the key points."
- "What did I write about onboarding last month?"
- "Show me my 5 most recent notes."
- "Create a note titled 'Standup 6/4' with today's three priorities."
Reading, searching, summarizing, and creating all happen through natural conversation.
Requirements
- macOS — fully supported and verified.
- Windows — best-effort. URLs launch through the registered scheme handler,
and the default DB path is guessed under
%APPDATA%\UpNote\. This path is not verified by the author — if reads fail, setUPNOTE_LENS_DB(see below). - Python 3.10+ (or just uv, which brings its own).
- UpNote desktop app installed.
Install & register
🤖 Let AI install it (easiest)
Give your MCP client (Claude, etc.) a link to
llms-install.md and it will run the steps and set everything
up for you.
🧑 Install it yourself
Pick the first option that fits what you already have. Each option includes how to register it in Claude.
Claude Desktop config file (referenced in each option):
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json- Windows:
%APPDATA%\Claude\claude_desktop_config.json
Option 1 — with uv (recommended)
Requires uv installed once
(brew install uv, or curl -LsSf https://astral.sh/uv/install.sh | sh).
After that there's no package install step — uvx downloads, caches, and
runs upnote-lens-mcp on demand.
Claude Code:
claude mcp add upnote-lens -- uvx upnote-lens-mcp
Claude Desktop:
{ "mcpServers": { "upnote-lens": { "command": "uvx", "args": ["upnote-lens-mcp"] } } }
Option 2 — with pip (if you don't have uv)
pip install upnote-lens-mcp
Claude Code:
claude mcp add upnote-lens -- upnote-lens-mcp
Claude Desktop:
{ "mcpServers": { "upnote-lens": { "command": "upnote-lens-mcp" } } }
Option 3 — from GitHub over HTTPS (if you don't have pip)
When you can't use PyPI/pip, install straight from the repo with uv. (No uv yet?
curl -LsSf https://astral.sh/uv/install.sh | sh.)
Claude Code:
claude mcp add upnote-lens -- uvx --from git+https://github.com/AwesomeHye/upnote-lens-mcp upnote-lens-mcp
Claude Desktop:
{ "mcpServers": { "upnote-lens": { "command": "uvx", "args": ["--from", "git+https://github.com/AwesomeHye/upnote-lens-mcp", "upnote-lens-mcp"] } } }
Tools
Read (queries the local DB → returns real text)
| Tool | Description |
|---|---|
search_notes(query, limit=20) |
Substring search over title/body. Returns id, title, updated time, snippet |
get_note(note_id, include_html=False) |
Full title + body text of a note (optionally raw HTML) |
list_recent(limit=20) |
Most recently updated notes |
list_notebooks() |
Notebooks with note counts and parent |
list_notes_in_notebook(notebook_id, limit=50) |
Notes inside a notebook |
list_tags() |
Tags with note counts |
list_notes_by_tag(tag_title, limit=50) |
Notes carrying a tag |
Write (upnote:// URL scheme)
| Tool | Description |
|---|---|
create_note(title, content, notebook?, markdown=True, new_window=False) |
Create a note. content is Markdown by default. notebook matches by name. Tags can't be set (see below) |
open_note(note_id, new_window=False) |
Open an existing note in the app |
open_notebook(notebook_id) |
Open a notebook in the app |
Tag limitation: UpNote's
note/newURL scheme has no tag parameter, and hashtags placed in the body stay as plain text rather than becoming real tags (they only convert to tags when typed in the editor). If you need tags, add them manually in the app after the note is created.
Override the DB path
If the database isn't in the default location (or you're on Windows), point at it with an environment variable:
UPNOTE_LENS_DB=/path/to/upnote.sqlite3
Default paths:
- macOS:
~/Library/Containers/com.getupnote.desktop/Data/Library/Application Support/UpNote/upnote.sqlite3 - Windows (best guess):
%APPDATA%\UpNote\upnote.sqlite3
License
MIT. See LICENSE for details. The URL-scheme formats and launch approach on the write side are adapted from chadthornton/upnote-mcp (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。