Anki Studio Tool Reference: Every MCP Tool Your AI Can Call

This page lists everything the Anki Studio MCP server offers your AI assistant: every tool it can call, grouped by feature. You don’t need to memorise any of it — your assistant discovers these automatically. This page is here for when you want to check exactly what exists.

At a glance

Documented against Studio 0.10.2. The server introduces itself to your AI client as ankimcp-studio, version 0.10.2, at https://studio.ankimcp.ai/mcp. Every tool is always available — there are no hidden or opt-in tools. Your account decides what the tools can see; it’s never a parameter.

What’s new in 0.10.2

• 9 new tools for reading what you have open: four pdf_reader_* tools for the PDF Reader and five video_reader_* tools for the Video Player.
• sandbox_push_card now copies media into Anki. Pictures and sounds from your Media Library that a card refers to are copied into Anki’s media folder before the note is added. Details are under the tool.
• Updated descriptions on sandbox_upsert_card, sandbox_patch_card, sandbox_get_cards, media_library_read_file and media_library_read_directory. They now tell your AI how to refer to Media Library files in a card, and that the library can hold PDFs and subtitle files. Their parameters didn’t change.

Nothing was renamed or removed.

Connection check

One tool to check that your AI is connected.

• ping — Health check. Optional message, which it echoes back. Returns the echo and your user ID. Changes nothing.

Knowledge maps

Build and edit your learning roadmaps. See Knowledge Roadmap.

Limits that apply across these tools: a map or folder title is 1–200 characters, a node label up to 200, a description up to 2,000, and a node’s data up to 8,192 characters. The only map type today is "roadmap". Free plan: 1 knowledge map.

• knowledge_map_create — Create a map. Requires type ("roadmap") and title; optional folderId to put it in a folder. Returns the new map, which starts with its root node. Free plan: 1 map.
• knowledge_map_list — List your maps. Returns a summary of each, without its nodes.
• knowledge_map_get — Get one map with all its nodes. Requires exactly one of mapId or title.
• knowledge_map_rename — Rename a map. Requires mapId and title. Returns the map.
• knowledge_map_delete — Delete a map. Requires mapId. Can’t be undone.
• knowledge_map_create_folder — Create a folder for maps. Requires title. Returns the folder.
• knowledge_map_rename_folder — Rename a folder. Requires folderId and title. Returns the folder.
• knowledge_map_delete_folder — Delete a folder. Requires folderId. Only an empty folder can be deleted.
• knowledge_map_list_folders — List your folders.
• knowledge_map_move_to_folder — Move a map into a folder. Requires mapId and folderId (null takes it out of any folder). Returns the map’s summary.
• knowledge_map_add_node — Add a node. Requires mapId, parentId and kind; optional label, description and data. Returns the map.
• knowledge_map_update_node — Change a node. Requires mapId and nodeId; optional label, description, data, and dataMode (merge, the default, or replace). Returns the map.
• knowledge_map_delete_node — Delete a node. Requires mapId and nodeId. Can’t be undone. Returns the map.
• knowledge_map_move_node — Move a node under a different parent. Requires mapId, nodeId and newParentId. Returns the map.
• knowledge_map_set_node_metrics — Set a node’s numbers by hand. Requires mapId, nodeId, and at least one of progress (0–100), cardCount or dueCount (0 or more). Returns the map.
• knowledge_map_group — Group two or more nodes. Requires mapId and nodeIds (at least 2); optional edgeMode (merged or linked). Returns the map.
• knowledge_map_ungroup — Undo a group. Requires mapId and groupId. Returns the map.
• knowledge_map_set_group_edge_mode — Change how a group’s lines are drawn. Requires mapId, groupId and edgeMode. Returns the map.
• knowledge_map_sync_stats — Fill in the map’s numbers from your real Anki. Requires mapId. Reads your Anki, so Anki must be running with the add-on. Overwrites the node numbers. Can take a while. Returns the map.

Media Library

Look up and label your stored pictures, sounds, PDFs and subtitle files. See Media Library.

There are no tools to upload, delete, move or rename files — you do that in the dashboard.

• media_library_read_directory — List everything under a folder, subfolders included. Optional path (default /, the top). Returns each folder and file; for files, the size, type, tags, description and mediaId. Changes nothing. Changed in 0.10.2: the description now mentions PDFs and subtitle files alongside images and sounds.
• media_library_read_file — Get one file’s details plus a fresh viewUrl that works for about 15 minutes. Requires url_or_path — a path, a bare mediaId, or a link the library gave out earlier (even an expired one). Optional include_content (default off): for an image or sound, your AI gets the file itself; for anything else, including PDFs and subtitles, it gets the details and a note saying why the file wasn’t included. Changed in 0.10.2: the description now covers PDFs and subtitle files, and tells your AI never to paste a viewUrl into a card — it stops working when it expires. Use <img src="ankimcp-media:<mediaId>"> or [sound:ankimcp-media:<mediaId>] instead.
• media_library_update_file — Change a file’s tags or description. Requires path, plus at least one of tags (up to 30, each 1–32 characters) or description (up to 2,000 characters). Writes to the library.

Sandbox

Draft cards and preview them before they reach Anki. See Sandbox.

Limits: up to 20 cards per sandbox. In the template, noteTypeName is 1–200 characters; fieldNames holds 1–50 names of up to 100 characters each; css and js are up to 256 KB each; qfmt and afmt (the front and back templates) up to 64 KB each. A card’s fields are up to 128 KB in total.

Media Library files go in a card as <img src="ankimcp-media:<mediaId>"> or [sound:ankimcp-media:<mediaId>] — in card fields only, not in the template or CSS.

• sandbox_get_session — Get the template, a short summary of every card, and the card count. Says so if there’s no sandbox yet. Saves whatever you’re typing in the dashboard editor first, so your AI sees your latest edits.
• sandbox_get_cards — Get full cards, either by ids (1–20) or by page with offset and limit (1–20) — not both. May return fewer cards than asked if they’re large. Saves your open edits first, like sandbox_get_session. Changed in 0.10.2: the description now says a card that refers to Media Library files must be sent with sandbox_push_card, which copies those files into Anki.
• sandbox_set_template — Replace the whole template. Requires noteTypeName, fieldNames, css, js, qfmt and afmt; optional createOnly. This is the only way a sandbox is created. Existing cards are kept.
• sandbox_update_template — Change part of the template. Requires at least one of noteTypeName, fieldNames, css, js, qfmt or afmt. Fails if there’s no sandbox yet.
• sandbox_patch_template — Find and replace inside one part of the template. Requires field (css, js, qfmt or afmt), oldStr and newStr. oldStr must match exactly once, and newStr must be different.
• sandbox_upsert_card — Create a card (leave out id) or replace a card’s fields. Requires fields; optional id. Returns the card’s id. Changed in 0.10.2: the description now explains how to show a Media Library file in a card, and that such a card must reach Anki through sandbox_push_card.
• sandbox_patch_card — Find and replace inside one field of one card. Requires id, field, oldStr and newStr (must be different). Changed in 0.10.2: the description now carries the same rule for Media Library files.
• sandbox_delete_card — Delete one card. Requires id. Can’t be undone.
• sandbox_push_card — Add one sandbox card to your Anki as a note. Requires cardId, deck_name and model_name; optional tags and allow_duplicate. Adds a note to your Anki (the add-on, or your Hosted Anki) and returns Anki’s reply. Changed in 0.10.2: before adding the note, it copies every Media Library picture and sound the card refers to into Anki’s media folder and points the card at the copies. Only png, jpg, webp, gif, mp3 and wav files are copied. If any file is another type, is missing, or fails to copy, no note is added. Files already copied stay in Anki and are reused next time. No add-on update is needed for this.
• sandbox_clear_session — Delete the whole sandbox: template and all cards. Can’t be undone. Safe to call when there’s nothing to clear.

Text to Speech

Turn text into spoken audio. See Text to Speech.

Credits per month: 500 on Free, 10,000 on Plus, 30,000 on Pro.

• tts_list_voices — List the available voices. Optional locale to narrow it down (for example, a language and its regional variants). Returns each voice’s name, language, gender and speaking styles.
• tts_generate — Make an mp3 and save it. Requires targetFolder ("" or "/" for the top of the library), text (1–5,000 characters), locale and voice. Optional fileName (1–128 characters), style, rate (−40 to 40) and pitch (−30 to 30). Returns the file’s path, the credits charged, and your remaining credits. Costs credits — Chinese, Japanese and Korean characters count double. Writes a file to your Media Library.
• tts_get_quota — Check your credits: your plan, monthly limit, used this month, and what’s left.

PDF Reader

Read the PDF you have open. See PDF Reader.

New in 0.10.2. These tools read the PDF open in the PDF Reader in your dashboard, in your browser. If nothing is open, they fail. They only read — they can’t change your PDF or your notes. You need a PDF project: 1 on Free, unlimited on Plus or Pro (up to 100). Projects are created in the dashboard, not by your AI.

• pdf_reader_status — New in 0.10.2. Describe the open PDF: the project, the file name, the current page and page count, the zoom, and the PDF’s own details (title, author and so on).
• pdf_reader_get_page — New in 0.10.2. Read one page. Requires page (1 or more). Optional format: "text" (the default), "image", or "pdf" for a one-page PDF. Optional annotations (image only, default off) draws your highlights and markup on the picture.
• pdf_reader_list_annotations — New in 0.10.2. List your highlights, notes, bookmarks, captures and other markup. Optional pages (from–to range), types (highlight, underline, strikeout, squiggly, ink, square, circle, freetext, note, bookmark, capture, other), fields to choose what comes back, limit (default 50, max 200) and offset.
• pdf_reader_get_capture — New in 0.10.2. Get the saved image of one capture, with its page, date, size and comment. Requires id.

Video Player

Read the video you have open. See Video Player.

New in 0.10.2. These tools read the video project open in the Video Player in your dashboard, in your browser. The video file itself never leaves your computer, so until you open it in the player your AI can’t get frames or subtitles — your bookmarks, drawings and captures still work. You need a video project: 1 on Free, unlimited on Plus or Pro (up to 100). Projects are created in the dashboard, not by your AI.

• video_reader_status — New in 0.10.2. Describe the open project: the file, whether the video is loaded, and if it is, its length, current time, whether it’s paused, speed and size. Also the subtitle file, if there is one.
• video_reader_get_frame — New in 0.10.2. Get one frame as a picture: the current one, the one at time (seconds), or the one showing an annotation (annotationId) — time and annotationId can’t be combined. With annotationId, optional view ("crop" or "overlay"). Without it, optional annotations draws your markup on the frame. With time or annotationId, it briefly seeks your video there and back.
• video_reader_get_subtitles — New in 0.10.2. Get subtitle lines with their start and end times. Optional time (from–to range), limit (default 100, max 500) and offset.
• video_reader_list_annotations — New in 0.10.2. List your bookmarks, spans, drawings, captures and other markup. Optional time (from–to range), types (bookmark, range, rect, freehand, text, arrow, ellipse, capture, other), fields, limit (default 50, max 200) and offset.
• video_reader_get_capture — New in 0.10.2. Get the saved image of one capture, with its time, date, size and comment. Requires id.

What your AI sees when something is off

Keeping the tool list fresh

Claude, ChatGPT and other clients remember the tool list from when you added Studio. When Studio gains tools, they won’t show up until you remove the Studio connector and add it again. Your existing connection keeps working in the meantime — nothing it already knows was removed or renamed.

Disclaimer: “Anki” is a registered trademark of Ankitects Pty Ltd. AnkiMCP is an independent, community-built project and is not affiliated with, endorsed by, or sponsored by Ankitects. MCP is an open standard originated by Anthropic; AnkiMCP is likewise not affiliated with or endorsed by Anthropic.