Ankimcp

AnkiMCP

[](https://github.com/matt-fff/ankimcp/actions/workflows/ci.yml) [](https://github.com/matt-fff/ankimcp/actions/workflows/test-matrix.yml) [](https://www.python.org/downloads/) [](https://github.com/psf/black) [](https://github.com/astral-sh/ruff)

Integrate your Anki decks with your choice of AI through the Model Context Protocol (MCP).

AnkiMCP is an Anki addon that exposes your Anki collection data (decks, notes, cards, and review statistics) via an MCP server, allowing AI assistants to help you study, create cards, and analyze your learning progress.

Features

• List all available decks with card counts
• Search notes using Anki's powerful search syntax
• Get detailed information about specific notes and cards
• View review statistics for decks or your entire collection
• Access card scheduling information and review history
• Create, update, and delete notes and decks
• Comprehensive permission system to control AI access
• Protect sensitive decks and notes from modification

Installation

As an Anki Addon

1. Download the addon from AnkiWeb or clone this repository
2. Copy the src/ankimcp folder to your Anki addons directory (or run

./install_addon.sh to copy files and vendor the mcp dependency)

1. Restart Anki
2. The MCP server will start automatically when you open your profile

Packaging for AnkiWeb

Use the helper script to generate a distributable .ankiaddon archive that matches the AnkiWeb sharing requirements:

python package_for_ankiweb.py --version 1.0.0

The resulting file is written to the dist/ directory (e.g. dist/ankimcp-1.0.0-YYYYmmdd-HHMMSS.ankiaddon). The script reads metadata from src/ankimcp/manifest.json, vendors the mcp runtime dependency into vendor/, and filters out common build artifacts so the archive is ready to upload to AnkiWeb. Use --skip-deps if you are packaging an already vendored source tree.

Development Setup

# Clone the repository
git clone https://github.com/yourusername/ankimcp.git
cd ankimcp

# Install dependencies with Rye
rye sync

# Run the test server (with mock data)
rye run python -m ankimcp

Usage

Once installed, the MCP server runs automatically when Anki is open. You can then connect to it from any MCP-compatible AI assistant.

Available MCP Tools

• list_decks - List all available Anki decks
• get_deck_info - Get detailed information about a specific deck
• search_notes - Search for notes using Anki's search syntax
• get_note - Get detailed information about a specific note
• get_cards_for_note - Get all cards associated with a note
• get_review_stats - Get review statistics for a deck or overall
• list_note_types - List all note types with their fields and templates
• create_deck - Create a new deck
• create_note_type - Create a new note type (card template)
• create_note - Create a new note in a deck
• update_note - Update an existing note's fields or tags
• update_deck - Update a deck's name or description
• delete_note - Delete a note and all its cards
• delete_deck - Delete a deck and all its cards
• get_permissions - View current permission settings

Permissions

AnkiMCP includes a comprehensive permission system to control AI access to your collection. You can:

• Set global read/write/delete permissions
• Use allowlists or denylists for deck access
• Protect specific decks from modification
• Restrict access based on note tags
• Control which note types can be created

See [PERMISSIONS.md](PERMISSIONS.md) for detailed configuration options.

Configuration

Edit the addon configuration in Anki (Tools → Add-ons → AnkiMCP → Config):

{
    "host": "localhost",
    "port": 4473,
    "permissions": {
        "global": {
            "read": true,
            "write": true,
            "delete": false
        },
        "mode": "denylist",
        "deck_permissions": {
            "denylist": ["Personal::*"]
        },
        "protected_decks": ["Default"]
    }
}

Testing & Acceptance

• Run rye run test to exercise the automated suite.
• COD-73 acceptance: packaged add-on bundles the mcp runtime dependency (via package_for_ankiweb.py or install_addon.sh) and the automated tests pass.

More Information

Visit https://ankimcp.com for more details and documentation.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: shivros
• Source: shivros/ankimcp
• License: MIT
• Homepage: https://ankimcp.com

Install and usage instructions live in the source repository linked above.