AI Whisper Voice Note Taker

Fetch one Whisprnotes note by ID with everything the app knows about it: complete text, tags, timestamps, word count, audio clips (with temporary playback URLs that expire after ~60 minutes), and whether the note has an active public share link. Use this after search returns a matching note. The ID should be a value returned by search, such as note:123.

fetch

List the authenticated user's tags with the number of notes on each. Use this to answer questions like "what topics do I have notes on?" or to discover valid tag names for search's tags filter.

list_tags

Search the authenticated user's Whisprnotes notes. Use this when the user asks about information that may be stored in their notes, or wants their notes used in the current task. Query should be ONE short natural-language phrase lifted from the user's request (e.g. "what to charge for the app"). The search is semantic; do not append keyword variants or synonyms — that hurts ranking. Pass an empty query string ("") to list the user's most recent notes (e.g. for "list all my notes"). The response's presentation field says what happened. "receipt": the match was clear-cut — each likely result carries its full note text in content; use that text to continue the user's task directly and do not enumerate the notes (the widget already confirms them). "refine": the match was ambiguous — the widget shows a list the user can narrow; reply with at most one short sentence. Optional filters: tags (note must have ALL listed tags; names are case-insensitive), created_after / created_before (ISO dates like 2026-07-01; created_after is inclusive, created_before exclusive). Pagination: limit (max 50) and offset; total_count in the response is the number of matches before pagination. Call fetch with a result id (e.g. note:123) to retrieve one full note.

search