Achriom

Get user context for adaptive conversation. Returns lifecycle stage and behavioral signals.

Get FULL details of an item FROM THE USER'S COLLECTION including AI analysis, user notes, rating, timeline, and all metadata. For albums: includes track list with durations. For box sets: lists ALL contained albums. Use this to answer specific questions about items the user owns.

Show the user's lists. With a name: that list's items in order. Without: all lists with counts.

List items by when they were ADDED to the library (newest first). Use for "show me the last N things I added", "my last 8 vinyl records", "what did I add recently". This is addition order — different from get_timeline, which orders by reading/watching dates. Returns title, creator, media_type, status, rating, and when each was added.

Get information about the current research scope.

Read a TV show's season-by-season watch progress and the next unwatched episode ("up next").

Get behavioral signals: theme repetition, consumption gaps, and recent activity.

Get collection statistics including progress, rating distribution, genre/theme breakdown, and timeline. If no media_type specified, returns combined stats.

Get the user's full Taste Portrait: their generated archetype, essence, core tension, preoccupations, the recommendation they don't own yet, and their blind spot. A compact version is already in your context; call this to go deeper when they ask about their portrait. Unlocks at 10 items; below that it returns how many more they need.

Get timeline showing items started and finished over time.

Get playable 30-second audio previews for tracks from an album IN THE USER'S LIBRARY.

Get the persistent profile built from past conversations. Shows taste patterns, key facts, cross-media connections, and preferences.

Log a completion or a re-watch/re-read/re-listen with a date. Use when the user mentions finishing something again ("watched Heat again last night", "reread Dune"). The first finish is tracked by status; this records repeats so they show in history and stats.

Search EXTERNAL databases (NOT the user's collection) to find items BEFORE adding them. Use search/get_details for items the user already owns. Data sources: Books=OpenLibrary, Movies=TMDB, Shows=TVDB (falls back to TMDB), Albums=Discogs, Anime=AniList, Games=IGDB. Use anime (not show) for Japanese animation.

Mark TV progress. Give season+episode to mark that episode; add through=true to mark everything up to it; give season only to mark the whole season; give neither to mark the whole show. Set watched=false to unmark.

Preview any album from Apple Music WITHOUT adding to library. Use this to sample before committing to add. Returns playable 30-second previews.

Pick random item(s) for serendipitous discovery.

Re-fetch metadata from external sources for an item whose data is incomplete. This re-runs the SAME external lookup, so it will return the same result as before: it does NOT fix a wrong match. If the item is the wrong film/show/album entirely, use edit_item with new_external_id to relink it, which is the only thing that changes which entry is fetched.

Read a specific section of an uploaded book by line numbers.

Remove an item from a list by title. The item stays in the library.

Save a lasting insight about the user to their persistent profile. Use when you discover meaningful patterns, preferences, personal connections to media, or cross-media themes.

Search the USER'S COLLECTION by title, creator, genre, or theme. Literal keyword match only - the query must actually appear in the item's title/creator/genre/theme text, or it returns nothing. Returns items they own with status, ratings, and for box sets, the list of contained albums. Use this to answer questions about what the user has. For generic, thematic, mood, or conceptual queries (e.g. "something about grief", "a film podcast") where the exact wording may not appear in any field, use `search_library` instead - it searches by meaning, not keyword.

Semantic search within an uploaded book (EPUB/PDF). Uses AI embeddings to find relevant passages.

Search past conversations with this user. Supports semantic search - finds conceptually related conversations, not just keyword matches.

Semantic search across the user's entire library by meaning, theme, or vibe. Searches every book/movie/album/show/anime/podcast/game as one corpus. Use for cross-media or thematic questions like "things about grief" or "noir mood", or any generic/conceptual query where the exact wording may not appear in the item's title/creator/genre text. For specific title/creator lookups, use the keyword `search` tool instead.

Search YouTube for relevant videos (interviews, trailers, analysis, music videos).

Set how a book was read: print, ebook, audiobook, or ereader.

Set the priority of an item in the collection. High = want this next, normal = standard backlog, low = speculative or distant.

Update how far the user is through an in-progress item. Books: current page. Shows: current season + episode. Anime: episodes watched.

ALWAYS use this tool when users say "show me", "open", "go to", "take me to", or "pull up" an item. This navigates them to the item's detail page. Works on ALL clients (web app, iOS app, Claude Desktop) - triggers navigation or returns clickable URL. Do NOT just describe the item when they want to SEE it.

Add or update personal notes for an item.

Set or update user rating. Supports half-stars: 0.5 to 5.0 in 0.5 increments (e.g. 3.5).

Update status. Books: unread, reading, finished, abandoned. Movies/shows/anime: unwatched, watching, watched, abandoned. Albums: unheard, listening, played, saved. Games: unplayed, playing, played, saved, on_hold, abandoned.