plaudlisted

# plaud Server: `plaud` (`mcp__plaud__*`). Reference: [references/plaud-api.md](references/plaud-api.md). ## Before any call - On `401`/`Not authenticated`, call `login`, wait for it to return, retry once; never ask the user for a token. - The Plaud API ignores `query`/`date_from`/`date_to`; the MCP `list_files` tool filters client-side over the newest 500 (`page`/`page_size` apply only unfiltered). - `get_transcript` is the largest payload: only where a mode names it, one page per call via `next_cursor` (scratchpad past three pages). - Relative dates resolve against today's date (reference table). - Show recordings as name, `YYYY-MM-DD`, `5m23s`, file ID. ## Quoting Paraphrase by default; copied wording goes in quotation marks with its source: `[MM:SS] Speaker` for a transcript line, `(Plaud summary)` for the note. A requested summary renders whole as a blockquote labeled `Plaud summary:`, synthesis kept separate. Facts come from the notes or transcript; unknowns are `?` or `TBD`. <example> - Weekly Sync (2026-09-01, 32m10s) — Launch moved to the 15th; Sam owns the migration checklist. Open question in the note: "who signs off on the DNS cutover" (Plaud summary). </example> ## browse — "what recordings do I have", "next page" `list_files(page=1, page_size=20)`; next page is `page+1`. Done: one page as a table, noting whether more exist (short page: no). ## find — "find the Weekly Sync", "the meeting from Monday", "the call about Q2" `list_files(query=, date_from=,