About Anki
anki-mcp is an MCP server that manages Anki flashcard collections through AnkiConnect, exposing 35 tools. It is aimed at learners who want to study, review statistics and bulk-edit cards from a chat client.
Key features
- 01Quizzes in chat with answers recorded in Anki
- 02Due counts, streak and review statistics
- 03Leech detection with per-card review history
- 04Bulk note creation that skips duplicates
- 05Label-based suspend and restore of card groups
- 06Optional Azure text-to-speech audio generation
Use cases
- 01Checking today's due cards and running a study session
- 02Finding frequently failed cards and rewriting them in bulk
- 03Turning an article into cloze cards in a chosen deck
MCP support
This project is an MCP server that exposes Anki study, card, note, deck and note type tools, plus resources and prompts, to MCP clients.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEanki-mcp
Study and manage your Anki collection by talking to Claude (or any MCP client). 35 tools over AnkiConnect: get quizzed in chat, see what is due and how your retention is going, find the cards you keep forgetting, create and clean up notes in bulk, and pause whole groups of cards until you want them back.
What you can ask
- "What do I have to study today?" → due counts per deck, streak, reviews this week.
- "Quiz me on 10 due cards from my Spanish deck." → one question at a time, graded, and recorded in Anki with the next interval.
- "Which cards do I keep getting wrong? Help me fix them." → leeches, their review history, and rewrites you approve.
- "Suspend everything tagged HSK5 until my exam, and make it easy to bring back." → named pause that restores exactly those cards, scheduling intact.
- "Turn this article into cloze cards in my Biology deck." → duplicate-checked bulk add.
- "Add audio to every card in this deck that doesn't have it." → Azure TTS in bulk.
- "Move the cards tagged
grammarto a new subdeck" / "rename the tagtodotoreview" / "spread my backlog over the next two weeks".
Bulk changes can be previewed with a dry run, deleting always needs a confirmation step, and when Anki is closed the assistant is told to ask you to open it instead of failing silently.
Install
You need Node.js 18+, Anki running, and the AnkiConnect add-on (code 2055492159).
Claude Desktop
Download anki-mcp.mcpb from the latest release and drag it into Settings → Extensions. That's it.
Claude Code
claude mcp add anki --scope user -- npx -y github:nietus/anki-mcp
Cursor and other MCP clients
Add this to the client's MCP configuration:
{
"mcpServers": {
"anki": {
"command": "npx",
"args": ["-y", "github:nietus/anki-mcp"]
}
}
}
On Windows, if the client cannot start npx directly, use "command": "cmd" with "args": ["/c", "npx", "-y", "github:nietus/anki-mcp"].
From source
git clone https://github.com/nietus/anki-mcp
cd anki-mcp
npm install
npm run build
Then point your client at node /path/to/anki-mcp/build/client.js. To build the Claude Desktop bundle yourself, run npm run pack:mcpb (output in dist/anki-mcp.mcpb).
Configuration
All environment variables are optional. Put them in a .env file in the project folder or pass them through your MCP client.
| Variable | Default | Purpose |
|---|---|---|
AZURE_API_KEY | – | Enables the audio tools (Azure Text-to-Speech). |
AZURE_REGION | eastus | Azure Speech region. |
ANKI_CONNECT_HOST | http://127.0.0.1 | AnkiConnect host. |
ANKI_CONNECT_PORT | 8765 | AnkiConnect port. |
ANKI_CONNECT_KEY | – | AnkiConnect API key, if you set one in the add-on config. |
Development
npm test # unit + in-memory MCP server tests (no Anki needed)
npm run inspector # try the tools interactively
Available Tools
Every tool is annotated as read-only, write or destructive, so clients like Claude Desktop can auto-approve reads and ask before destructive changes. Errors (including "Anki is not running") come back as readable tool results.
Most bulk tools select cards or notes the same way: deckName (subdecks included), tags (matches any), query (any Anki search, ANDed), or explicit cardIds / noteIds. Write tools accept dryRun: true to preview the change first.
Studying
| Tool | What it does |
|---|---|
get_study_overview | Today's new/learning/review counts per deck, cards reviewed today, streak, 7/30-day totals. |
get_due_cards / get_new_cards | num cards to study (optionally from one deckName), with the interval each button would give. |
answer_cards | Record answers (cardId, ease 1-4) and get each card's new interval. (Previously update_cards.) |
find_leeches | Weakest cards (leech tag / many lapses), ranked by lapses and ease. |
get_card_history | Full review log of up to 20 cards. |
get_retention_stats | True retention (young/mature), reviews per day and time spent for a deck over N days. |
undo | Undo the last action in Anki. |
Selecting and scheduling cards
| Tool | What it does |
|---|---|
find_cards | Paginated search (query, limit, offset); pick the sections to return with include: question, answer, fields, stats. |
suspend_cards / unsuspend_cards | Deactivate/reactivate cards, e.g. all cards with tag X in deck Y. Scheduling is preserved. Pass a label when suspending (notes get the tag paused::<label>) and unsuspend_cards { label } later resumes exactly that group, leaving cards suspended for other reasons alone. |
move_cards | Move cards to targetDeck (created if missing). |
reschedule_cards | set_due_date (days: 0, 1!, 3-7), forget (reset to new) or relearn. |
Creating and editing notes
| Tool | What it does |
|---|---|
add_card | Create one note. |
add_bulk | Create many notes; duplicates/invalid notes are skipped and reported with the reason. |
add_cloze_cards | Create cloze notes from text plus terms to hide, or text with {{c1::...}} markup. |
update_note_fields / bulk_update_notes | Edit fields of existing notes. |
find_and_replace | Plain or regex replace across notes, optionally limited to some fields. |
manage_tags | list (with a filter: tag counts), add, remove, rename, clear_unused. |
add_image | Add an image to a field from a url, local path or base64 data. |
delete_notes | Two-step delete: the first call previews and returns a confirmToken, the second call (with the token) deletes. |
Audio (requires AZURE_API_KEY)
| Tool | What it does |
|---|---|
add_card_with_audio | Create a note with audio generated from sourceField into audioField. |
update_card_with_audio | Generate audio for one existing note. |
bulk_generate_audio | Generate audio for many notes (by default only those with an empty audio field), limit per call. |
Supported languages: en, es, fr, de, it, ja, ko, pt, pt-PT, ru, zh, ar, nl, hi, tr, pl, sv, fi, da, no, cs, hu, el, he, th, vi, id, ms, ro.
Decks and note types
| Tool | What it does |
|---|---|
get_deck_names / create_deck | List or create decks. |
get_deck_model_info | Which note types a deck uses (call before adding cards to it). |
get_model_names / get_model_details | List note types / fields, templates and CSS of one. |
edit_note_type_field | add, remove, rename or reposition a field. |
update_note_type_templates / update_note_type_styling | Replace templates or CSS. |
create_model | Create a note type. |
sync | Sync with AnkiWeb. |
Prompts
| Prompt | What it does |
|---|---|
quiz_me | Study session: one question at a time, graded and recorded in Anki. |
make_cards | Turn a text into atomic flashcards (previewed before adding). |
leech_review | Find the cards you keep forgetting and fix them. |
Resources
anki://collection/overview– study overview.anki://search/isdue,anki://search/isnew,anki://search/deckcurrent– up to 50 cards each.anki://deck/{deckName}/overview– due counts for one deck.anki://model/{modelName}– fields, templates and CSS of a note type.
License
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.
