Deckhand

Give your AI assistant a careful set of hands inside Anki.

You already talk to Claude, Codex, or another AI assistant. Deckhand lets that assistant work directly inside your live Anki collection — your real decks, your note types, your review history — instead of guessing from a pasted CSV or a stale export. It sees what you see, proposes changes in plain language, and nothing touches your collection until you approve it.

What your assistant can do with it

• Create — Drop in lecture slides, a PDF, or pasted notes and get well-formed, cited cards that fit your existing decks and note types. Every card is checked against a card-design rubric and previewed before anything is added.
• Check — Get your decks reviewed the way engineers review code: vague prompts, overloaded cloze deletions, missing context, ambiguous answers — each finding comes with a proposed fix.
• Fix — "I keep failing these." Deckhand diagnoses leeches and weak cards, names the actual failure mode, and shows an exact before/after for every repair.
• Organize — Untangle tags, hunt down duplicates, restructure decks, and get filtered-deck suggestions — with the exact scope confirmed before any bulk change.

Why this instead of a card-generator website

Deckhand is a bridge that lets your assistant work through Anki's own runtime and APIs while Anki is open in front of you:

• You stay in control. Preview before apply, always.
• Backups before anything big. Your assistant is instructed to create an Anki backup before bulk edits, imports, template changes, or scheduling changes.
• No database hacking. All work goes through Anki's APIs — never direct edits to your collection database or media folder.
• Context-aware, not generic. Your assistant inspects your current deck, related cards, and note types before proposing changes.

Private by design

Deckhand runs entirely on your computer. The bridge binds to loopback (127.0.0.1) by default and never uploads your collection anywhere. Want an extra local guard? Enable an access token in Deckhand's settings.

AI assistants can still make bad suggestions. Review changes before applying them, and keep backups before large operations.

Install

1. Install this add-on from AnkiWeb, then restart Anki.
2. Follow the Deckhand welcome dialog.
3. Missed the dialog? Use Deckhand -> Onboarding or Deckhand -> Management in Anki's menu bar.

For Claude Desktop, there is a one-click Deckhand.mcpb package on GitHub Releases. Claude Code, Codex, and other MCP clients get tailored setup instructions inside the add-on.

Requirements

• Anki Desktop with modern add-on support.
• An MCP-capable AI assistant, such as Claude Desktop, Claude Code, Codex, or any Streamable HTTP MCP client.
• No network access needed for Deckhand itself — the bridge is fully local.

Source, license, and support

Source code:

https://github.com/bentoware/deckhand

Release notes:

https://github.com/bentoware/deckhand/releases/

Issues and bug reports:

https://github.com/bentoware/deckhand/issues

License:

AGPL-3.0-or-later

Deckhand is provided as-is, without warranty, to the extent permitted by applicable law. The authors and contributors are not liable for lost data, damaged decks, incorrect study material, or other damages arising from use of the software or AI-assisted workflows.