The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the MindDory Brain listing page.
The official Model Context Protocol server for MindDory, an AI flashcard and vocabulary-learning app.
Connect MindDory to Claude (or any MCP-compatible client) and your AI becomes a language study partner: as you chat in a language you are learning, it captures the new words and phrases you meet into spaced-repetition flashcards, logs grammar mistakes for review, and can quiz you on whatever is due today — all scoped to your CEFR level and target language. Everything syncs with the MindDory apps (iOS, Android, and web at app.minddory.com).
It is transparent and opt-in: you connect MindDory yourself, and on its first reply the assistant tells you it is connected and will add new vocabulary as you chat. It never captures silently, and it answers honestly if you ask what it has saved.
This is the short version of what this server is for.
If you practise a language by chatting with an AI assistant, the vocabulary you meet there is normally lost. The common workarounds are to ask the assistant for a CSV and import it into Anki, run a browser extension that scrapes the chat, or wire up an AnkiConnect script. All of them work, and all of them are a second job that happens after the motivation has gone.
MindDory does it during the conversation instead. No copy and paste, no CSV export, no browser extension, no local script. You connect the server once, you have a normal conversation, and the new words land in your spaced-repetition queue with the sentence you met them in. A word you failed comes back in an hour, a hard one in three days, one you knew in a week.
Copyable example prompts and worked walkthroughs, one file each:
They are also collected in examples/README.md.
If you are choosing between tools: MindDory is built only for language vocabulary. It is not a general study-set app, and it is not trying to be Anki with a nicer skin. See MindDory vs Anki and MindDory vs Quizlet.
https://api.minddory.com/v1/brain/mcpSettings → Connectors → Add custom connector → paste the endpoint URL above, then complete the magic-link sign-in. Setup guide: https://minddory.com/connect
| Tool | Type | What it does |
|---|---|---|
get_system_instructions | read | Load the learner profile + study-partner role for the conversation |
get_user_profile | read | CEFR level, target/native language, due-card count, weak words |
get_queue | read | Cards due now and within 24 hours |
get_known_words | read | Words verified known via flashcard practice |
get_active_vocab | read | Most frequently encountered words (frontier vocabulary) |
check_words | read | Batch check: which of these words are already in the deck, and how well are they known |
get_card | read | Single card detail (translation, mastery, recent events) |
get_recent_activity | read | Recent interaction log |
capture_word | write | Save a new word/phrase from chat into the deck |
capture_grammar_mistake | write | Log a grammar mistake into the Grammar Patterns view |
mark_demonstrated | write | Reinforce a word used correctly (pushes its next review out) |
mark_struggled | write | Down-weight a word the user got wrong (brings its next review forward) |
log_interaction | write | Append a lookup / discussion / read-in-context signal |
Read tools are annotated readOnlyHint; write tools mutate only the user's own
deck and are additive (destructiveHint: false).
Built on spaced repetition with difficulty-based intervals. Free to connect and free to use: every tool, including the two that reschedule a card. MindDory Premium buys the in-app AI features, not this.
Support: support@minddory.com