Check that the local OpenWhispr app is running and its CLI bridge is reachable. Returns the bridge host and port, the handshake file path and the app version. Call this first when another tool fails with bridge_not_running, bridge_unreachable or unauthorized — the port and token change on every app restart.

List notes newest first (by updated_at), filtered by note_type and/or folder_id. Returns summaries only — no note body, no transcript. Pages come from one snapshot: pass next_cursor back verbatim, keeping the same filters. A cursor that outlived its snapshot fails with snapshot_expired; list again without a cursor.

Read one note by id: title, folder, timestamps and the full content body. The transcript is never included — it reaches 240 KB — so a meeting note reports transcript_segment_count and a hint pointing at get_note_transcript. Set include_enhanced to also receive the AI-cleaned version of the text.

Full-text (FTS5 prefix AND) search over note titles, bodies and AI-enhanced text. Every word becomes a required prefix term, so all of them must appear. There is no semantic search and no relevance score. Returns note summaries with matched_in and a snippet; use list_notes to browse and get_note for a full body.

Read a note's transcript. format=segments pages diarized segments with times relative to the recording start (offset/limit, speaker/source filters); format=text renders "[mm:ss] speaker: …" lines; format=speakers gives per-speaker totals only — use it first on long recordings. A note with no transcript returns an empty result, not an error.

Create a note in the local OpenWhispr app. Without folder_id the app files it into its own default folder — see the notice field in the reply. note_type is fixed at creation: update_note cannot change it. Returns the stored note.

Change a note's title, content or folder in the local OpenWhispr app. note_type, transcript and enhanced_content are deliberately not writable. Rewriting content leaves any existing enhanced_content in place and the app keeps treating it as current; the reply warns when that happens.

Delete a note from the local OpenWhispr app. There is no confirmation step and no undo on this path, so confirm with the user before calling it. The bridge answers 204 without saying whether the id existed.

List the note folders of the local OpenWhispr app, with their ids, names and default flags. Use it to find the folder_id that list_notes, create_note and update_note take. The bridge returns every folder at once — there is no limit and no paging here.

Create a note folder in the local OpenWhispr app. Names must be unique: a duplicate comes back as folder_name_conflict listing the folders that exist. Creating a folder is the only folder change the bridge allows.

List the OpenWhispr dictation history (newest first): text, provider, model, status and audio duration. These rows are NOT note transcripts and cannot be linked to a note — use get_note_transcript for those. Discarded and deleted dictations are invisible here.

Read one dictation from the history by id, with its full text, provider, model, status and any error. Ids come from list_transcriptions. A dictation is not linked to any note — use get_note_transcript for a note transcript.

List the custom dictionary of the local OpenWhispr app — the words the transcriber is told to spell a particular way (product names, jargon, names of people). Returns the words themselves plus the shape the app stored them in.

Add or remove words in the custom dictionary of the local OpenWhispr app. Words are trimmed and de-duplicated, and case is significant. Returns what was sent plus the dictionary as it reads back afterwards.

Summarise what is stored in the local OpenWhispr app: how many notes, folders, transcriptions and dictionary words there are, split by type, folder and month, with word, character and audio totals. Counts come from a capped read, and the reply lists exactly what it cannot see.