upnote-lens-mcp
# 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](https://github.com/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, set `UPNOTE_LENS_DB` (see below).
- Python 3.10+ (or just [uv](https://docs.astral.sh/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`](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](https://docs.astral.sh/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:
```bash
claude mcp add upnote-lens -- uvx upnote-lens-mcp
```
Claude Desktop:
```json
{ "mcpServers": { "upnote-lens": { "command": "uvx", "args": ["upnote-lens-mcp"] } } }
```
#### Option 2 — with pip (if you don't have uv)
```bash
pip install upnote-lens-mcp
```
Claude Code:
```bash
claude mcp add upnote-lens -- upnote-lens-mcp
```
Claude Desktop:
```json
{ "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:
```bash
claude mcp add upnote-lens -- uvx --from git+https://github.com/AwesomeHye/upnote-lens-mcp upnote-lens-mcp
```
Claude Desktop:
```json
{ "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/new` URL 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](LICENSE) for details.
The URL-scheme formats and launch approach on the write side are adapted from
chadthornton/upnote-mcp (MIT).
TDQS
Scored across 10 tools
Each tool targets a distinct resource and action: creation, retrieval by id, listing by various criteria (notebooks, tags, recent), searching, and opening in the app. There is no ambiguity because the operations are clearly separated by resource type (note, notebook, tag) and action verb.
All tool names follow a consistent verb_noun pattern (e.g., create_note, get_note, list_notebooks, search_notes). The naming uses lowercase and underscores uniformly, making it predictable and easy for an agent to infer function from the name.
With 10 tools, the server provides a well-scoped set for note management. Each tool has a clear purpose and none feel redundant. This count is within the ideal range (3-15) and balances coverage without overwhelming the agent.
The server covers creation, reading, listing, searching, and opening notes/notebooks, but lacks update and delete operations. Additionally, there is no tool to manage tags (add/remove). These are notable gaps that may limit an agent's ability to perform full lifecycle management.