anki-mcp
Enables reading and writing to your local Anki collection through AnkiConnect, allowing creation, update, and search of notes and decks from MCP clients like Claude Desktop.
README
anki-mcp
An MCP server that exposes your local Anki collection to MCP clients such as Claude Desktop and VS Code, through the AnkiConnect add-on.
Note fields are HTML and are passed through verbatim, so cards can be generated with cloze deletions, bold/italic, and highlight.js-styled code blocks.
Tools
Read
anki_check_connection— verify Anki is running and AnkiConnect is reachableanki_list_decks— every deck (::denotes nesting)anki_list_models— every note typeanki_get_model_fields— a note type's field names, in template orderanki_find_notes— note ids matching an Anki search queryanki_get_notes_info— full note details: model, tags, fields, cards
Write
anki_create_deck— create a deck, including missing parentsanki_add_note— add one note (fields are HTML, sent unescaped)anki_add_notes— add a batch sharing one deck and note typeanki_update_note_fields— overwrite fields of an existing noteanki_sync— sync the collection with AnkiWeb
Requirements
- uv
- Anki, running — the server talks to the live application, not the collection file
- The AnkiConnect add-on: Anki → Tools → Add-ons → Get Add-ons → code
2055492159, then restart Anki
Setup
uv sync
cp .env.example .env
.env (all optional — the defaults work for a standard local Anki):
ANKI_CONNECT_URL=http://localhost:8765
ANKI_TIMEOUT=30
Run it:
uv run anki-mcp
Check it can reach Anki:
uv run python scripts/smoke_client.py
Use with Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"anki": {
"command": "uv",
"args": ["run", "--directory", "/absolute/path/to/anki-mcp", "anki-mcp"]
}
}
}
No secrets to pass — AnkiConnect is unauthenticated and local.
Remote (HTTP) hosting
Set MCP_TRANSPORT=http to serve over streamable-http (endpoint /mcp):
MCP_TRANSPORT=http HOST=0.0.0.0 PORT=8000 uv run anki-mcp
This is of limited use here. AnkiConnect binds to localhost on the machine where Anki runs, so a remotely hosted instance of this server has nothing to connect to — every tool would fail. The transport toggle exists for structural parity with a normal MCP server; making it genuinely useful would need a tunnel back to the desktop (a WebSocket relay or similar), which is out of scope. Run it locally over stdio.
Conventions
- Fields are HTML. Values in
fieldsare sent to Anki byte for byte — no escaping, no sanitizing."<b>x</b>"renders as a bold x; to show a literal<, send<. - Cloze deletions use
{{c1::...}}and require a cloze note type; the syntax is inert on"Basic". - Field names are case- and space-sensitive (
"Back Extra"). Check them withanki_get_model_fieldsbefore adding notes. - Duplicates:
anki_add_notefails on a duplicate first field;anki_add_notesskips the offending note and reports its index. Both acceptallow_duplicate=True. anki_update_note_fieldssilently does nothing if the note is open in Anki's Browse window — Anki refuses the edit but reports success. Close the browser and retry.- Tool output is Markdown by default; pass
response_format="json"for the raw payload (and for untruncated field HTML). - Anki must stay open, and it stops serving requests while a modal dialog is up.
Development
uv sync --extra dev
uv run pytest # tests (HTTP mocked, no network, no real Anki)
uv run ruff check . # lint
uv run ruff check --fix .
uv run fastmcp inspect src/anki_mcp/server.py:mcp
Security
AnkiConnect has no authentication: anything that can reach port 8765 can read and modify your collection. That is acceptable because it listens on localhost only — do not expose that port, and do not run this server on an untrusted machine. .env is gitignored, though it holds no secrets by default.
This server can create, overwrite, and sync notes. Back up your collection (Anki → File → Export) before letting an agent make bulk changes.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.