zettelkasten-mcp

MCP.Pizza Chef: entanglr

Fourteen commands let your assistant write single-idea notes, connect them with labelled relationships such as supports, contradicts or refines, tag them, search them, list the most connected notes and surface the ones nothing points to. Notes live as ordinary text files on your own machine alongside a small index, so you keep everything and no key or account is needed. Deleting a note erases its file outright, with no bin and no undo. The project has had no changes since April 2025.

Unmaintained · no changes since April 2025
Notes
Writing

Use This MCP server To

Turn a reading session into a set of one-idea notes Link a new idea to the older note it argues against Find which of my notes nothing else links to See which notes sit at the centre of my thinking Search my notes by tag or by wording Build an outline note that organises a topic

README

Zettelkasten MCP Server

A Model Context Protocol (MCP) server that implements the Zettelkasten knowledge management methodology, allowing you to create, link, explore and synthesize atomic notes through Claude and other MCP-compatible clients.

What is Zettelkasten?

The Zettelkasten method is a knowledge management system developed by German sociologist Niklas Luhmann, who used it to produce over 70 books and hundreds of articles. It consists of three core principles:

  1. Atomicity: Each note contains exactly one idea, making it a discrete unit of knowledge
  2. Connectivity: Notes are linked together to create a network of knowledge, with meaningful relationships between ideas
  3. Emergence: As the network grows, new patterns and insights emerge that weren't obvious when the individual notes were created

What makes the Zettelkasten approach powerful is how it enables exploration in multiple ways:

  • Vertical exploration: dive deeper into specific topics by following connections within a subject area.
  • Horizontal exploration: discover unexpected relationships between different fields by traversing links that cross domains.

This structure invites serendipitous discoveries as you follow trails of thought from note to note, all while keeping each piece of information easily accessible through its unique identifier. Luhmann called his system his "second brain" or "communication partner" - this digital implementation aims to provide similar benefits through modern technology.

Features

  • Create atomic notes with unique timestamp-based IDs
  • Link notes bidirectionally to build a knowledge graph
  • Tag notes for categorical organization
  • Search notes by content, tags, or links
  • Use markdown format for human readability and editing
  • Integrate with Claude through MCP for AI-assisted knowledge management
  • Dual storage architecture (see below)
  • Synchronous operation model for simplified architecture

Examples

Note Types

The Zettelkasten MCP server supports different types of notes:

Type Handle Description
Fleeting notes fleeting Quick, temporary notes for capturing ideas
Literature notes literature Notes from reading material
Permanent notes permanent Well-formulated, evergreen notes
Structure notes structure Index or outline notes that organize other notes
Hub notes hub Entry points to the Zettelkasten on key topics

Link Types

The Zettelkasten MCP server uses a comprehensive semantic linking system that creates meaningful connections between notes. Each link type represents a specific relationship, allowing for a rich, multi-dimensional knowledge graph.

Primary Link Type Inverse Link Type Relationship Description
reference reference Simple reference to related information (symmetric relationship)
extends extended_by One note builds upon or develops concepts from another
refines refined_by One note clarifies or improves upon another
contradicts contradicted_by One note presents opposing views to another
questions questioned_by One note poses questions about another
supports supported_by One note provides evidence for another
related related Generic relationship (symmetric relationship)

Prompting

To ensure maximum effectiveness, we recommend using a system prompt ("project instructions"), project knowledge, and an appropriate chat prompt when asking the LLM to process information, or explore or synthesize your Zettelkasten notes. The docs directory in this repository contains the necessary files to get you started:

System prompts

Pick one:

Project knowledge

For end users:

Chat Prompts

Project knowledge (dev)

For developers and contributors:

NB: Optionally include the source code with a tool like repomix.

Storage Architecture

This system uses a dual storage approach:

  1. Markdown Files: All notes are stored as human-readable Markdown files with YAML frontmatter for metadata. These files are the source of truth and can be:

    • Edited directly in any text editor
    • Placed under version control (Git, etc.)
    • Backed up using standard file backup procedures
    • Shared or transferred like any other text files
  2. SQLite Database: Functions as an indexing layer that:

    • Facilitates efficient querying and search operations
    • Enables Claude to quickly traverse the knowledge graph
    • Maintains relationship information for faster link traversal
    • Is automatically rebuilt from Markdown files when needed

If you edit Markdown files directly outside the system, you'll need to run the zk_rebuild_index tool to update the database. The database itself can be deleted at any time - it will be regenerated from your Markdown files.

Installation

# Clone the repository
git clone https://github.com/entanglr/zettelkasten-mcp.git
cd zettelkasten-mcp

# Create a virtual environment with uv
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
uv add "mcp[cli]"

# Install dev dependencies
uv sync --all-extras

Configuration

Create a .env file in the project root by copying the example:

cp .env.example .env

Then edit the file to configure your connection parameters.

Usage

Starting the Server

python -m zettelkasten_mcp.main

Or with explicit configuration:

python -m zettelkasten_mcp.main --notes-dir ./data/notes --database-path ./data/db/zettelkasten.db

Connecting to Claude Desktop

Add the following configuration to your Claude Desktop:

{
  "mcpServers": {
    "zettelkasten": {
      "command": "/absolute/path/to/zettelkasten-mcp/.venv/bin/python",
      "args": [
        "-m",
        "zettelkasten_mcp.main"
      ],
      "env": {
        "ZETTELKASTEN_NOTES_DIR": "/absolute/path/to/zettelkasten-mcp/data/notes",
        "ZETTELKASTEN_DATABASE_PATH": "/absolute/path/to/zettelkasten-mcp/data/db/zettelkasten.db",
        "ZETTELKASTEN_LOG_LEVEL": "INFO"
      }
    }
  }
}

Available MCP Tools

All tools have been prefixed with zk_ for better organization:

Tool Description
zk_create_note Create a new note with a title, content, and optional tags
zk_get_note Retrieve a specific note by ID or title
zk_update_note Update an existing note's content or metadata
zk_delete_note Delete a note
zk_create_link Create links between notes
zk_remove_link Remove links between notes
zk_search_notes Search for notes by content, tags, or links
zk_get_linked_notes Find notes linked to a specific note
zk_get_all_tags List all tags in the system
zk_find_similar_notes Find notes similar to a given note
zk_find_central_notes Find notes with the most connections
zk_find_orphaned_notes Find notes with no connections
zk_list_notes_by_date List notes by creation/update date
zk_rebuild_index Rebuild the database index from Markdown files

Project Structure

zettelkasten-mcp/
├── src/
│   └── zettelkasten_mcp/
│       ├── models/       # Data models
│       ├── storage/      # Storage layer
│       ├── services/     # Business logic
│       └── server/       # MCP server implementation
├── data/
│   ├── notes/            # Note storage (Markdown files)
│   └── db/               # Database for indexing
├── tests/                # Test suite
├── .env.example          # Environment variable template
└── README.md

Tests

Comprehensive test suite for Zettelkasten MCP covering all layers of the application from models to the MCP server implementation.

How to Run the Tests

From the project root directory, run:

Using pytest directly
python -m pytest -v tests/
Using UV
uv run pytest -v tests/
With coverage report
uv run pytest --cov=zettelkasten_mcp --cov-report=term-missing tests/
Running a specific test file
uv run pytest -v tests/test_models.py
Running a specific test class
uv run pytest -v tests/test_models.py::TestNoteModel
Running a specific test function
uv run pytest -v tests/test_models.py::TestNoteModel::test_note_validation

Tests Directory Structure

tests/
├── conftest.py - Common fixtures for all tests
├── test_integration.py - Integration tests for the entire system
├── test_mcp_server.py - Tests for MCP server tools
├── test_models.py - Tests for data models
├── test_note_repository.py - Tests for note repository
├── test_search_service.py - Tests for search service
├── test_semantic_links.py - Tests for semantic linking
└── test_zettel_service.py - Tests for zettel service

Important Notice

⚠️ USE AT YOUR OWN RISK: This software is experimental and provided as-is without warranty of any kind. While efforts have been made to ensure data integrity, it may contain bugs that could potentially lead to data loss or corruption. Always back up your notes regularly and use caution when testing with important information.

Credit Where Credit's Due

This MCP server was crafted with the assistance of Claude, who helped organize the atomic thoughts of this project into a coherent knowledge graph. Much like a good Zettelkasten system, Claude connected the dots between ideas that might otherwise have remained isolated. Unlike Luhmann's paper-based system, however, Claude didn't require 90,000 index cards to be effective.

License

MIT License

zettelkasten-mcp FAQ

Do I need a key or an account?
No. Everything runs on your own computer. Your notes are plain text files in a folder you choose, and nothing is sent to an outside service by the server itself.
Can I use this to organise my reading notes?
Yes. It has separate kinds of note for quick thoughts, notes from reading, polished permanent notes, outlines and topic entry points, and your assistant can file each one properly as you go.
Can it delete my notes?
Yes, and permanently. The delete command removes the note's file from disk immediately, with no confirmation and no recycle bin. Keep the folder in version control or a backup before you let an assistant tidy up.
Does 'find similar notes' understand what the notes say?
No, and this surprises people. Similarity is worked out from shared tags and shared links only, never the wording. Two notes on the same subject with no tags or links in common score zero.
How hard is the setup?
Hard. You download the code, install Python tooling, create an environment and then point your assistant at a local command. There is no installer and no hosted version.
Can I still edit notes in my own text editor?
Yes, the files are the real thing and the index is just a helper. If you edit them outside the tool, run the rebuild command afterwards so searching stays accurate.
Which apps does it work in?
Claude Desktop is what the instructions cover, and any other MCP-compatible app that can start a local command will work the same way.
Is it still being worked on?
It has been quiet since April 2025, with no changes for well over a year. It still runs, but do not expect fixes or new features.