Krisp

Fetch one or more documents by their IDs. Returns the parts selected by `include` -- notes, summary, key points, action items, tags, a recording link, or the transcript -- for meetings, and agenda details for upcoming meetings. When a request spans several meetings, default to notes, summary and action items (or key_points): they carry the decisions, follow-ups and topics discussed, and exist even when search results show no pre-computed summary yet. The transcript is heavy -- tens of thousands of characters per meeting -- so fetch it one meeting at a time, and only when the user explicitly asks for the full transcript or verbatim wording. Any request covering more than 10 meetings must use light parts only (see Limits). ## Supported ID Types - `meeting_id` from search_meetings or list_action_items -- returns the meeting's requested parts - `agenda_id` from list_upcoming_meetings or search_meetings -- returns agenda details ## Limits The per-request id cap depends on how much content is requested: - Up to 50 ids when every requested part is light: `summary`, `key_points`, `action_items`, `tags`, `recording`. - Up to 10 ids when `include` contains `notes` or `transcript`, or when `include` is omitted (whole document). Each of those returns a full body. To digest many meetings at once (e.g. "summarize all sales calls this month"), pass `include: ["key_points"]` or `["summary"]` (or `["action_items"]`, etc.) with up to 50 ids in a single call. Ask for `notes` or `transcript` only on the meetings that need the full text, in batches of 10. The whole reply is capped at 120,000 characters shared across the returned documents. Every requested id is always present; when documents are large, each is cut with a marker stating how many characters were dropped and how to fetch the rest. ## ID Format Document IDs are 32 lowercase hexadecimal characters (UUID without dashes). ## Action items When you need this meeting's action items WITH their ids (e.g. to mark one done), request `include: ["action_items"]`. Each item then carries its `action_item_id` (in markdown as `action_item_id: <id>`; in JSON as the `id` field) — pass that to complete_action_item. Omitting `include` returns the plain document text, which may not show ids. For a whole meeting's items, `list_action_items` with `meeting_id` is usually simpler. ## Response Format Returns an array of results, one for each requested ID: `[{id: string, document: string | object | null}]` `document` is a markdown string (format=markdown) or a JSON object (format=json). Documents that weren't found will have `document: null`.

get_multiple_documents

Returns the current user's display preferences including name, timezone, and company. These are user-configured values that provide context for formatting responses. ## Response Fields - name: The user's display name - timezone: The user's timezone - company: The user's company domain

get_user_preferences

Start importing an audio or video recording as a new Krisp meeting. Returns a one-time pre-signed upload URL. This tool does NOT upload the file itself — hand the returned url to the client/host application, which uploads the file bytes to it with an HTTP PUT (Content-Type = the file's real type, e.g. video/mp4 or audio/mpeg). That upload alone creates the meeting and starts transcription; there is no separate "complete" step. ## Flow 1. Call this tool → get { import_id, url }. 2. The client/host PUTs the file bytes to url (out-of-band; Content-Type = the file's real type). 3. Poll get_import_status with import_id until status is "ready", then use meeting_id. ## Timing (important for correct operation) Start imports SEQUENTIALLY: call this tool, WAIT for its result, then start the next one. Only one import start per user may be in flight at a time. If you fire several at once (before the previous call returns), the overlapping ones fail with an "Action is still in process" error — that is NOT a failure of the recording; just retry that one after the in-flight start returns. There is no fixed cooldown: once a call returns, you can immediately start the next. All parameters are optional, but sending every value you have is highly recommended — they give the meeting better metadata (title, language, duration). If you don't have a value, just omit it; that's fine. Only send values you're confident about: a value that fails validation is rejected immediately with an error, whereas an omitted value is accepted without complaint. The recording's type is taken from the uploaded file's Content-Type. The uploaded file must be a supported audio/video recording (mp3, mpeg, aac, mp4, m4a, wav, wma, quicktime, wmv) and no larger than 1 GB. Empty, oversized, or non-audio/unsupported uploads are rejected — get_import_status then returns status "failed" with an error reason, and no meeting is created.

import_recording

List action items (tasks/follow-ups) from meetings. Returns action items with their completion status, assignee, and source meeting information. ## Response Fields - id: Action item ID — pass this as action_item_id to complete_action_item to mark it done/not done - title: The action item description - completed: Whether the task is done - assignee: Name of assigned person (or null) - due_date: Due date if set (ISO 8601) - meeting_id: Source meeting 32-char UUID - meeting_name: Source meeting title - meeting_date: When the meeting occurred - total: Total count of matching action items (for pagination) - notes_unavailable, notice: Set with meeting_id when that meeting has notes but no structured action items section; read its note body with get_multiple_documents and include: ["notes"] ## Usage Action item titles are NOT unique — multiple items can share the same text. Always show each item's action_item_id when presenting items to the user, and use the id (never the title) to act on a specific item via complete_action_item. If the user asks to change a specific item without giving an id, list the relevant items with their ids and ask which one first.

list_action_items

List activities from the Activity Center. Returns recent notifications and updates including action item assignments, meeting summaries, and other activity feed items. ## Response Fields - source_type: Type of activity (e.g., "action_item_assignment_activity") - url: Direct link to the activity in Krisp - title: Activity title in markdown format - description: Activity description in markdown format - created_at: When the activity was created (ISO 8601) - total: Total count of activities - unread: Count of unread activities

list_activities

List or search the user's meeting folders. Folders group related meetings together. Use this to: - Resolve a folder ID to its name and metadata - Find a folder by name - See how many meetings/members a folder has ## Response Fields - folder_id: Unique folder ID - name: Folder name - description: Folder description (or null) - icon: The folder's icon: { type: "icon", name, color } for a named app icon, { type: "emoji", emoji } for an emoji, or null - visibility: "private" or "teamspace" - my_role: The user's role on the folder ("owner" / "collaborator" / "viewer") - meeting_count: Number of meetings in the folder - member_count: Number of members - shared_with_others: true when people other than the user can see the folder (a team folder, or a private folder shared with others) ## Tips - To list the meetings inside a folder, use get_folder with the folder_id. - To file meetings into a folder or take them out, use add_to_folder / remove_from_folder with the folder_id. To make a new private folder, use create_folder. - A folder created moments ago may not appear here yet; use the folder_id create_folder returned.

list_folders

List and search tags. Use this tool to discover tag names and IDs. **Important:** When the user wants to filter meetings by tag, call this tool FIRST to resolve tag names to IDs, then pass those IDs to the search_meetings tool's `tags` parameter. ## Modes ### 1. List / Search Tags (default) Returns all tags, optionally filtered by name. Supports autocomplete-style prefix matching. ### 2. Most Used Tags Set `most_used: true` to get tags ranked by how many meetings they appear in. Useful for discovering frequently-used tags. ## Response Fields - id: Tag ID (32-char hex string) - use this with search_meetings `tags` parameter - name: Tag display name - color: Tag color code - creator_name: Name of the person who created the tag (when available) - usage_count: Number of meetings with this tag (only in most_used mode) - total: Total count of matching tags (for pagination)

list_tags

List upcoming calendar meetings for the next few days. Returns scheduled meetings with title, time, participants, organizer, and conferencing details. ## Response Fields - agenda_id: 32-char UUID for the meeting agenda - title: Meeting title - start: Start time (ISO 8601) - end: End time (ISO 8601) - status: Calendar event status (confirmed, tentative, cancelled) - organizer: Organizer name - is_external: Whether the organizer is external to the workspace - is_recurring: Whether this is a recurring meeting - participants: List of attendee names - conferencing_app: Video conferencing app (zoom, meet, teams, etc.) - conferencing_url: Link to join the meeting

list_upcoming_meetings

Reassign speakers in a meeting's transcript — one speaker or all of them in a single call. Both the transcript and the summary update (same as renaming a speaker in the app).

bulk_update_speakers

Remove meetings from one of the user's folders, in ONE call (up to 50 meetings). - `folder_id`: from list_folders or get_folder. - `meeting_ids`: every meeting to take out, all in this one call. Never loop one call per meeting. - The meetings themselves are not deleted; they are only taken out of the folder. - The user must be the folder's owner or a collaborator; on a team folder, a team admin counts as an owner and other team members cannot remove meetings. - The owner can remove any meeting; a collaborator only meetings they own. - A meeting the user cannot open is reported as not found. - A meeting that is not in the folder counts as done. The result lists what was removed, what was not in the folder, and what could not be removed and why. Report all of it; do not retry the whole batch.

remove_from_folder

Search meetings by text query or retrieve by ID. Returns meeting metadata, summaries, action items, and key points. ## Two Modes ### 1. ID Lookup Use the `id` parameter for direct retrieval. Returns meeting metadata (name, date, attendees, speakers — including diarized speakers with no assigned person, see `speakers` below), summaries, action items, and recurring meeting history. Does not include the full transcript. ### 2. Text Search Use the `search` parameter to find meetings by name, content discussed, action items, key points, or attendee names. ## Response Fields - meeting_id: Unique 32-char UUID (always included) - name: Meeting title - date: Meeting date/time (start) - end_time: Meeting end time (calendar end, or start + duration when unavailable) - duration_seconds: Recording length in seconds - url: Direct link to meeting - is_recurring: Boolean for recurring meetings - attendees: List of attendee names - speakers: List of people who spoke. A speaker the transcript detected but nobody has assigned to a person appears under the same label the transcript uses — "Speaker 1", "Speaker 2", ... — in this same list, so the count reflects everyone who spoke, not just the named ones. Pass such a label to bulk_update_speakers to assign it. - tags: List of tag names on the meeting - transcript: { status } - transcript availability status - agenda: { agenda_id } - 32-char UUID for the meeting agenda - meeting_notes: { detailed_summary, key_points, action_items }; also carries notes_unavailable: true and a notice when the notes have no structured action items section or nothing structured could be returned - past_meeting_occurrences: For recurring meetings, previous instances with their meeting_id ## Data Scope This tool returns pre-computed summaries and metadata only. meeting_notes is absent when a meeting has no notes yet, or none of the requested parts. Notes written from a template without structured sections (e.g. 1:1, Outline or a custom template) may have no structured action items, or nothing structured at all: those meetings get meeting_notes.notes_unavailable: true and a notice, next to any summary that was returned — this is not a meeting without notes; read the note body with get_multiple_documents and include: ["notes"]. Full transcripts and complete meeting content are not included in search results.

search_meetings