The title clearly communicates the tool’s action and subject. Keep it concise.
Description98/100
List the authenticated user's tasks. Use status to choose which set: "needs-action" (the default) for open tasks, "completed" for finished ones. One call returns one set or the other, never both — issue two calls if you need both. AI-suggested (pending/triaging) and dismissed (rejected) tasks are never returned by this tool. Filters: priority, tagId, updatedAfter. Only one of priority/tagId can be used per request. Priority order is high, medium, normal, low. "normal" (the default) sits between medium and low. Returns tasks with their titles, due dates, priority, tags, and more, most recently modified first, capped at 50 per call. If truncated is true the cap was reached and more matched: narrow with updatedAfter before counting or summarizing. Large pages may omit optional fields; those names appear in elidedFields when that happens.
The description explains the capability, when to use it, and what to expect.
Inputs90/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
status
Which tasks to list. "needs-action" (the default) returns open tasks; "completed" returns finished tasks instead of open ones, not alongside them. Pair "completed" with updatedAfter to find recently finished work — recency is by last-modified time, not completion time, so a task finished earlier but edited recently also matches. Say so when reporting.
Clear enough to supply this argument.
priority
Filter by priority. Priority order is high, medium, normal, low. "normal" (the default) sits between medium and low. Cannot be combined with tagId in one request.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
tagId
Filter by tag ID (returns tasks that have this tag)
Clear enough to supply this argument.
updatedAfter
ISO 8601 local date-time or calendar date; only return tasks updated after this time. Do not include Z or a timezone offset — the value is read in ianaTimeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
ianaTimeZone
IANA time zone (e.g. Europe/Zurich) used to interpret dates and to express returned timestamps. Defaults to the user's configured time zone.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_tasks",
"title": "List Tasks",
"description": "List the authenticated user's tasks. Use status to choose which set: \"needs-action\" (the default) for open tasks, \"completed\" for finished ones. One call returns one set or the other, never both — issue two calls if you need both. AI-suggested (pending/triaging) and dismissed (rejected) tasks are never returned by this tool. Filters: priority, tagId, updatedAfter. Only one of priority/tagId can be used per request. Priority order is high, medium, normal, low. \"normal\" (the default) sits between medium and low. Returns tasks with their titles, due dates, priority, tags, and more, most recently modified first, capped at 50 per call. If truncated is true the cap was reached and more matched: narrow with updatedAfter before counting or summarizing. Large pages may omit optional fields; those names appear in elidedFields when that happens.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"tagId": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Filter by tag ID (returns tasks that have this tag)"
},
"status": {
"anyOf": [
{
"enum": [
"needs-action",
"completed"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "Which tasks to list. \"needs-action\" (the default) returns open tasks; \"completed\" returns finished tasks instead of open ones, not alongside them. Pair \"completed\" with updatedAfter to find recently finished work — recency is by last-modified time, not completion time, so a task finished earlier but edited recently also matches. Say so when reporting."
},
"priority": {
"anyOf": [
{
"enum": [
"high",
"medium",
"normal",
"low"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "Filter by priority. Priority order is high, medium, normal, low. \"normal\" (the default) sits between medium and low. Cannot be combined with tagId in one request."
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
],
"description": "IANA time zone (e.g. Europe/Zurich) used to interpret dates and to express returned timestamps. Defaults to the user's configured time zone."
},
"updatedAfter": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "ISO 8601 local date-time without timezone suffix (e.g. 2025-06-15T09:00:00) or calendar date (e.g. 2025-06-15). Do not include Z or a timezone offset — the zone comes from ianaTimeZone, which defaults to the user's configured time zone."
},
{
"type": "null"
}
],
"description": "ISO 8601 local date-time or calendar date; only return tasks updated after this time. Do not include Z or a timezone offset — the value is read in ianaTimeZone."
}
}
},
"annotations": {
"title": "List Tasks",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description93/100
Retrieve one or more tasks by their IDs. Returns an array in input order. Each entry is either a task object (type: 'task') or an error object ({ id, error }) for IDs that could not be fetched.
The description explains the capability, when to use it, and what to expect.
Inputs50/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
ids
Array of task IDs to retrieve (obtained from list_tasks)
Clear enough to supply this argument.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_tasks",
"title": "Get Tasks",
"description": "Retrieve one or more tasks by their IDs. Returns an array in input order. Each entry is either a task object (type: 'task') or an error object ({ id, error }) for IDs that could not be fetched.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"ids"
],
"properties": {
"ids": {
"type": "array",
"items": {
"type": "string",
"minLength": 1
},
"minItems": 1,
"description": "Array of task IDs to retrieve (obtained from list_tasks)"
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
}
}
},
"annotations": {
"title": "Get Tasks",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description71/100
Create a new task. At minimum a title is required. Optional fields: description, due date, estimated duration, priority, tags. Priority order is high, medium, normal, low. "normal" (the default) sits between medium and low.
The description explains the capability, when to use it, and what to expect.
Inputs25/100
2 of 8 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
title
Not provided
Missing description. Explain what this value means and how to supply it.
description
Not provided
Missing description. Explain what this value means and how to supply it.
estimatedDurationMinutes
Not provided
Missing description. Explain what this value means and how to supply it.
priority
Not provided
Missing description. Explain what this value means and how to supply it.
tags
Tag IDs to assign (obtained from list_tags, or from create_tag for a new tag). Tag names are not accepted; unknown ids are rejected.
Clear enough to supply this argument.
due
Not provided
Missing description. Explain what this value means and how to supply it.
earliestStartDate
Not provided
Missing description. Explain what this value means and how to supply it.
source
URL linking to the source of this task (e.g. Slack thread, email, web page)
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description73/100
Update an existing task. Only provided fields are changed.
Explain mutations, irreversible effects, or prerequisites that matter before calling this tool.
Inputs90/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
id
ID of the task to update (obtained from list_tasks)
Clear enough to supply this argument.
title
Updated task title. Null is ignored — titles cannot be cleared.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
description
Updated description. null is no change; clear it with clearFields.
Clear enough to supply this argument.
estimatedDurationMinutes
Estimated duration in minutes. null is no change; clear it with clearFields.
Clear enough to supply this argument.
priority
Task priority. Priority order is high, medium, normal, low. "normal" (the default) sits between medium and low. "normal" clears it to the default. null is no change.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
tags
Tag IDs (obtained from list_tags, or from create_tag for a new tag) — replaces the full tag list. null is no change; clear all tags with clearFields. Tag names are not accepted; unknown ids are rejected.
Clear enough to supply this argument.
due
Due date as YYYY-MM-DD (e.g. 2025-06-15). null is no change; clear it with clearFields.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
earliestStartDate
Earliest start date as YYYY-MM-DD (e.g. 2025-06-15). null is no change; clear it with clearFields.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
source
URL linking to the source of this task (e.g. Slack thread, email, web page). null is no change; remove it with clearFields.
Clear enough to supply this argument.
clearFields
Fields to clear: description, estimatedDurationMinutes, priority (to the default), tags (all), due, earliestStartDate, source. Do not also send a value for a cleared field.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "close_task",
"title": "Close Task",
"description": "Mark a task as completed. If the task is recurring, completing it automatically advances the task to its next occurrence.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"minLength": 1,
"description": "ID of the task to mark as completed (obtained from list_tasks)"
}
}
},
"annotations": {
"title": "Close Task",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description92/100
Reopen a previously completed task, marking the whole task as not completed. This does not target a single occurrence of a recurring task and does not undo any recurrence advance caused by a prior close — it only clears the task's completed state.
The description explains the capability, when to use it, and what to expect.
Inputs97/100
Input documentation gives an agent enough context to populate the arguments.
id
ID of the completed task to reopen (obtained from list_tasks)
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "reopen_task",
"title": "Reopen Task",
"description": "Reopen a previously completed task, marking the whole task as not completed. This does not target a single occurrence of a recurring task and does not undo any recurrence advance caused by a prior close — it only clears the task's completed state.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"minLength": 1,
"description": "ID of the completed task to reopen (obtained from list_tasks)"
}
}
},
"annotations": {
"title": "Reopen Task",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description94/100
List the tags (categories) of the authenticated user, most recently modified first, 50 per page. When nextCursor is present, calling again with cursor fetches the next page; a tag is only known to be absent once no nextCursor remains. Tag ids can be used with list_tasks to filter by tag, with create_task/update_task to assign tags, and with update_tag/delete_tag. create_tag creates a tag that does not exist yet.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
cursor
Opaque cursor from a previous list_tags response's nextCursor. Omit to start from the first page.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_tags",
"title": "List Tags",
"description": "List the tags (categories) of the authenticated user, most recently modified first, 50 per page. When nextCursor is present, calling again with cursor fetches the next page; a tag is only known to be absent once no nextCursor remains. Tag ids can be used with list_tasks to filter by tag, with create_task/update_task to assign tags, and with update_tag/delete_tag. create_tag creates a tag that does not exist yet.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"cursor": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Opaque cursor from a previous list_tags response's nextCursor. Omit to start from the first page."
}
}
},
"annotations": {
"title": "List Tags",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description97/100
Create a new tag (category) for the authenticated user and return it. The returned id is what create_task / update_task tags accept. Fails if a tag with the same name already exists (list_tags shows existing tags).
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
name
Tag name. Must not match an existing tag of the user (case-sensitive).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
color
Hex color in #rrggbb form (e.g. #ff8800). A random light color is assigned when omitted.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "create_tag",
"title": "Create Tag",
"description": "Create a new tag (category) for the authenticated user and return it. The returned id is what create_task / update_task tags accept. Fails if a tag with the same name already exists (list_tags shows existing tags).",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"name"
],
"properties": {
"name": {
"type": "string",
"minLength": 1,
"description": "Tag name. Must not match an existing tag of the user (case-sensitive)."
},
"color": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Hex color in #rrggbb form (e.g. #ff8800). A random light color is assigned when omitted."
}
}
},
"annotations": {
"title": "Create Tag",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description97/100
List the user's calendars (name, account, access rights, sharing). By default returns calendars the user has not hidden in the client app. Set includeHidden to return all calendars, including calendars they hid or disabled. Each calendar has role (primary | contextual) and optional isHidden. Role meanings (also returned as roleExplanation): primary — Calendar whose events block availability and should be considered for scheduling decisions. contextual — Calendar used as contextual information, not as the user's own schedule. Its events can be shown/read, but they do not mean the user is busy: exclude them from availability, free/busy calculations and open-slot searches. isSharedWithMe / ownedBy mark calendars shared with the user.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
includeHidden
Include all calendars, including calendars the user has hidden or disabled in the client app. Leave this false unless you need calendars that are normally out of view. Defaults to false (visible calendars only).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_calendars",
"title": "List Calendars",
"description": "List the user's calendars (name, account, access rights, sharing). By default returns calendars the user has not hidden in the client app. Set includeHidden to return all calendars, including calendars they hid or disabled. Each calendar has role (primary | contextual) and optional isHidden. Role meanings (also returned as roleExplanation): primary — Calendar whose events block availability and should be considered for scheduling decisions. contextual — Calendar used as contextual information, not as the user's own schedule. Its events can be shown/read, but they do not mean the user is busy: exclude them from availability, free/busy calculations and open-slot searches. isSharedWithMe / ownedBy mark calendars shared with the user.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"includeHidden": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "Include all calendars, including calendars the user has hidden or disabled in the client app. Leave this false unless you need calendars that are normally out of view. Defaults to false (visible calendars only)."
}
}
},
"annotations": {
"title": "List Calendars",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description94/100
Update the authenticated user's Kai booking-link defaults and return the full saved configuration. Only provided properties change: null, blank and empty values are no change. Availability patches merge by weekday; close a weekday with closeDays. Timezone changes preserve wall-clock hours.
The description explains the capability, when to use it, and what to expect.
Inputs86/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
meetingTitle
New meeting title. Omitted or null leaves it unchanged.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
durationMinutes
Allowed meeting durations in minutes. The first value is the default. Omitted or null leaves them unchanged.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
timezone
IANA timezone for availability. A timezone-only change keeps the same wall-clock hours. Omitted or null leaves it unchanged.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description98/100
List calendar events within a date range. Request only the date range you need. Maximum range is 1 year. Timed event start/end are returned as wall-clock times in timeZone (or the user's configured time zone), matching the client app calendar display. All-day event end is exclusive (a one-day event starting 2026-09-15 has end 2026-09-16). Large lists may omit optional fields under token pressure; omitted names appear in elidedFields. Pass includeFields only when strictly needed. When calendars are omitted, each event has calendarRole: only primary events are the user's own schedule; contextual events are context and must not count as the user being busy. Cancelled events are omitted. Invites the account owner declined are omitted by default (includeDeclined=true includes them; they are marked isDeclined). When calendars are omitted, automatically queries the user's visible calendars across all connected accounts. When calendars are explicitly provided, only events:read scope is required. partialFailures lists what could not be read: a calendar (calendarId, plus calendarName when calendars were omitted) when only some calendars of an account failed, or the whole account otherwise.
The description explains the capability, when to use it, and what to expect.
Inputs61/100
5 of 7 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
start
Start of the query range in ISO 8601 local date-time (e.g. 2025-01-01T00:00:00). Do not include Z or a timezone offset — the range is read in timeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
end
End of the range in ISO 8601 local date-time (e.g. 2025-01-31T23:59:59). Do not include Z or a timezone offset — the range is read in timeZone. Request only the date range you need. Maximum range is 1 year.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
timeZone
Not provided
Missing description. Explain what this value means and how to supply it.
calendars › anyOf › 0 › items › accountId
Calendar account ID (obtained from list_calendars)
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
calendars › anyOf › 0 › items › calendarIds
Not provided
Missing description. Explain what this value means and how to supply it.
includeDeclined
Include events the account owner declined. Defaults to false (client app calendar default). Declined invites often remain status=confirmed on Google — they are not cancelled. When included, those declined events have isDeclined=true / userRSVPStatus=declined.
Clear enough to supply this argument.
includeFields
Compressible event fields to keep even when the list is large (calendarName, attendees, description, privacy, recurrence, color). description is omitted from list_events unless listed here. Use only when strictly needed for this call — each field increases response size. Requested fields may make the response exceed its normal token budget. For details on a few specific events, prefer get_events instead.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_events",
"title": "List Events",
"description": "List calendar events within a date range. Request only the date range you need. Maximum range is 1 year. Timed event start/end are returned as wall-clock times in timeZone (or the user's configured time zone), matching the client app calendar display. All-day event end is exclusive (a one-day event starting 2026-09-15 has end 2026-09-16). Large lists may omit optional fields under token pressure; omitted names appear in elidedFields. Pass includeFields only when strictly needed. When calendars are omitted, each event has calendarRole: only primary events are the user's own schedule; contextual events are context and must not count as the user being busy. Cancelled events are omitted. Invites the account owner declined are omitted by default (includeDeclined=true includes them; they are marked isDeclined). When calendars are omitted, automatically queries the user's visible calendars across all connected accounts. When calendars are explicitly provided, only events:read scope is required. partialFailures lists what could not be read: a calendar (calendarId, plus calendarName when calendars were omitted) when only some calendars of an account failed, or the whole account otherwise.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"start",
"end"
],
"properties": {
"end": {
"type": "string",
"minLength": 1,
"description": "End of the range in ISO 8601 local date-time (e.g. 2025-01-31T23:59:59). Do not include Z or a timezone offset — the range is read in timeZone. Request only the date range you need. Maximum range is 1 year."
},
"start": {
"type": "string",
"minLength": 1,
"description": "Start of the query range in ISO 8601 local date-time (e.g. 2025-01-01T00:00:00). Do not include Z or a timezone offset — the range is read in timeZone."
},
"timeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
},
"calendars": {
"anyOf": [
{
"type": "array",
"items": {
"type": "object",
"required": [
"accountId",
"calendarIds"
],
"properties": {
"accountId": {
"type": "string",
"minLength": 1,
"description": "Calendar account ID (obtained from list_calendars)"
},
"calendarIds": {
"type": "array",
"items": {
"type": "string",
"minLength": 1,
"description": "Calendar ID from that account"
},
"minItems": 1
}
}
},
"minItems": 1
},
{
"type": "null"
}
],
"description": "Calendars to query. When omitted, automatically resolves the user's visible calendars across all accounts (requires calendars:read scope)."
},
"includeFields": {
"anyOf": [
{
"type": "array",
"items": {
"enum": [
"calendarName",
"attendees",
"description",
"privacy",
"recurrence",
"color"
],
"type": "string"
}
},
{
"type": "null"
}
],
"description": "Compressible event fields to keep even when the list is large (calendarName, attendees, description, privacy, recurrence, color). description is omitted from list_events unless listed here. Use only when strictly needed for this call — each field increases response size. Requested fields may make the response exceed its normal token budget. For details on a few specific events, prefer get_events instead."
},
"includeDeclined": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "Include events the account owner declined. Defaults to false (client app calendar default). Declined invites often remain status=confirmed on Google — they are not cancelled. When included, those declined events have isDeclined=true / userRSVPStatus=declined."
}
}
},
"annotations": {
"title": "List Events",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description94/100
Retrieve one or more calendar events by their IDs. Timed event start/end are returned as wall-clock times in timeZone (or the user's configured time zone), matching the client app calendar display. All-day event end is exclusive (a one-day event starting 2026-09-15 has end 2026-09-16). Returns full event details for found events (including cancelled or declined ones, with isCancelled / isDeclined) and error entries for events that could not be retrieved.
The description explains the capability, when to use it, and what to expect.
Inputs50/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
ids
Array of event IDs to retrieve (obtained from list_events, or a meeting session's calendarEventId or calendarSeriesId)
Clear enough to supply this argument.
timeZone
Not provided
Missing description. Explain what this value means and how to supply it.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_events",
"title": "Get Events",
"description": "Retrieve one or more calendar events by their IDs. Timed event start/end are returned as wall-clock times in timeZone (or the user's configured time zone), matching the client app calendar display. All-day event end is exclusive (a one-day event starting 2026-09-15 has end 2026-09-16). Returns full event details for found events (including cancelled or declined ones, with isCancelled / isDeclined) and error entries for events that could not be retrieved.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"ids"
],
"properties": {
"ids": {
"type": "array",
"items": {
"type": "string",
"minLength": 1
},
"minItems": 1,
"description": "Array of event IDs to retrieve (obtained from list_events, or a meeting session's calendarEventId or calendarSeriesId)"
},
"timeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
}
}
},
"annotations": {
"title": "Get Events",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description87/100
Create a new calendar event. Requires account ID, calendar ID, title, start time, duration, and isAllDay. For timed events, timeZone defaults to the user's time zone when omitted or null. For all-day events, duration is a whole number of calendar days (P1D is one day, P2D is two); the exclusive end is start plus duration (a one-day event starting 2026-09-15 has end 2026-09-16). An all-day start can be a date (2026-09-15) or local midnight (2026-09-15T00:00:00). Optionally add attendees — this sends calendar invitations via the provider. By default, a video conferencing link (Google Meet or Microsoft Teams) is added based on the calendar provider.
The description explains the capability, when to use it, and what to expect.
Inputs57/100
9 of 15 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
accountId
Account ID to create the event in (obtained from list_calendars)
Clear enough to supply this argument.
title
Not provided
Missing description. Explain what this value means and how to supply it.
description
Not provided
Missing description. Explain what this value means and how to supply it.
calendarId
Calendar ID to create the event in (obtained from list_calendars)
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
start
Start time in ISO 8601 local date-time format (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — this is the wall-clock start in timeZone, which defaults to the user's time zone. An all-day event also accepts a date (e.g. 2026-09-15), read as 2026-09-15T00:00:00.
Clear enough to supply this argument.
duration
Event duration in ISO 8601 duration format (e.g. PT1H for 1 hour, PT30M for 30 minutes, P1D for one all-day calendar day). All-day events use whole-day durations: P1D is one day, P2D is two consecutive days. The exclusive end date is start plus duration (a one-day event on 2026-09-15 has end 2026-09-16).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
timeZone
Not provided
Missing description. Explain what this value means and how to supply it.
isAllDay
Whether this is an all-day event. When true, timeZone should be null. All-day duration is counted in whole calendar days with an exclusive end (P1D covers only the start date; list_events / get_events return end as the next day).
Clear enough to supply this argument.
freeBusyStatus
Not provided
Missing description. Explain what this value means and how to supply it.
privacy
Not provided
Missing description. Explain what this value means and how to supply it.
location
Not provided
Missing description. Explain what this value means and how to supply it.
attendees › anyOf › 0 › items › email
Attendee email address
Clear enough to supply this argument.
attendees › anyOf › 0 › items › name
Attendee display name
Clear enough to supply this argument.
taskId
Link this event to a task by ID (obtained from list_tasks). Creates a scheduled time block for the task.
Clear enough to supply this argument.
addConferencingLink
Add a video conferencing link (Google Meet for Google calendars, Microsoft Teams for Outlook) to the event. Defaults to true; set to false to create an event without a conferencing link.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "create_event",
"title": "Create Event",
"description": "Create a new calendar event. Requires account ID, calendar ID, title, start time, duration, and isAllDay. For timed events, timeZone defaults to the user's time zone when omitted or null. For all-day events, duration is a whole number of calendar days (P1D is one day, P2D is two); the exclusive end is start plus duration (a one-day event starting 2026-09-15 has end 2026-09-16). An all-day start can be a date (2026-09-15) or local midnight (2026-09-15T00:00:00). Optionally add attendees — this sends calendar invitations via the provider. By default, a video conferencing link (Google Meet or Microsoft Teams) is added based on the calendar provider.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"accountId",
"title",
"calendarId",
"start",
"duration",
"isAllDay"
],
"properties": {
"start": {
"type": "string",
"minLength": 1,
"description": "Start time in ISO 8601 local date-time format (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — this is the wall-clock start in timeZone, which defaults to the user's time zone. An all-day event also accepts a date (e.g. 2026-09-15), read as 2026-09-15T00:00:00."
},
"title": {
"type": "string",
"minLength": 1
},
"taskId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"description": "Link this event to a task by ID (obtained from list_tasks). Creates a scheduled time block for the task."
},
"privacy": {
"anyOf": [
{
"enum": [
"public",
"private",
"secret"
],
"type": "string",
"description": "Event visibility: public (visible to everyone), private (details hidden), or secret (not visible)"
},
{
"type": "null"
}
]
},
"duration": {
"type": "string",
"minLength": 1,
"description": "Event duration in ISO 8601 duration format (e.g. PT1H for 1 hour, PT30M for 30 minutes, P1D for one all-day calendar day). All-day events use whole-day durations: P1D is one day, P2D is two consecutive days. The exclusive end date is start plus duration (a one-day event on 2026-09-15 has end 2026-09-16)."
},
"isAllDay": {
"type": "boolean",
"description": "Whether this is an all-day event. When true, timeZone should be null. All-day duration is counted in whole calendar days with an exclusive end (P1D covers only the start date; list_events / get_events return end as the next day)."
},
"location": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"timeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone for start/end wall-clock times (e.g. Europe/Zurich). For timed events this is the request/user display zone — not the event's original storage zone."
},
{
"type": "null"
}
]
},
"accountId": {
"type": "string",
"minLength": 1,
"description": "Account ID to create the event in (obtained from list_calendars)"
},
"attendees": {
"anyOf": [
{
"type": "array",
"items": {
"type": "object",
"required": [
"email"
],
"properties": {
"name": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Attendee display name"
},
"email": {
"type": "string",
"minLength": 1,
"description": "Attendee email address"
}
}
}
},
{
"type": "null"
}
],
"description": "Event attendees. Adding attendees sends calendar invitations via the provider."
},
"calendarId": {
"type": "string",
"minLength": 1,
"description": "Calendar ID to create the event in (obtained from list_calendars)"
},
"description": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
]
},
"freeBusyStatus": {
"anyOf": [
{
"enum": [
"free",
"busy"
],
"type": "string",
"description": "Whether the event blocks availability (busy) or not (free)"
},
{
"type": "null"
}
]
},
"addConferencingLink": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "Add a video conferencing link (Google Meet for Google calendars, Microsoft Teams for Outlook) to the event. Defaults to true; set to false to create an event without a conferencing link."
}
}
},
"annotations": {
"title": "Create Event",
"readOnlyHint": false,
"openWorldHint": true,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description77/100
Update an existing calendar event. Only provided fields are changed. Timing updates interpret start as display wall-clock times in timing.timeZone, or the user's time zone when timing.timeZone is null; the event's original storage timezone is preserved. For all-day events, timing.duration is a whole number of calendar days (P1D is one day, P2D is two). list_events and get_events return an exclusive all-day end: a one-day event starting 2026-09-15 has end 2026-09-16, matching duration P1D. An all-day timing.start can be a date (2026-09-15) or local midnight (2026-09-15T00:00:00).
The description explains the capability, when to use it, and what to expect.
Inputs81/100
16 of 17 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
id
ID of the event to update (obtained from list_events)
Clear enough to supply this argument.
accountId
Account ID that owns the event (obtained from list_calendars)
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
calendarId
Calendar ID that contains the event (obtained from list_calendars)
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
title
Not provided
Missing description. Explain what this value means and how to supply it.
timing › anyOf › 0 › start
Updated start as wall-clock local date-time (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — this is the display wall-clock in timing.timeZone (defaults to the user's time zone). The event's storage timezone is preserved server-side. When timing.isAllDay is true, a date (e.g. 2026-09-15) is also accepted, read as 2026-09-15T00:00:00.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
timing › anyOf › 0 › duration
Updated duration in ISO 8601 duration format (e.g. PT1H for 1 hour, P1D for one all-day calendar day). All-day events use whole-day durations: P1D is one day, P2D is two consecutive days. The exclusive end date is start plus duration (a one-day event on 2026-09-15 has end 2026-09-16).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
timing › anyOf › 0 › timeZone
Display IANA time zone for interpreting start (e.g. Europe/Zurich). Send null to use the user's time zone. Does not change the event's storage timezone.
Clear enough to supply this argument.
timing › anyOf › 0 › isAllDay
Whether this is an all-day event. All-day duration is counted in whole calendar days with an exclusive end (P1D covers only the start date; list_events / get_events return end as the next day).
Clear enough to supply this argument.
description
Updated event description in plain text or HTML. null is no change; clear it with clearFields.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
freeBusyStatus
Whether the event blocks availability (busy) or not (free). null is no change; reset it with clearFields.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
privacy
Event visibility: public, private, or secret. null is no change; reset it with clearFields.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
location
Event location name. null is no change; clear it with clearFields.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
clearFields
Fields to clear: description, location, or freeBusyStatus and privacy (to the default). Do not also send a value for a cleared field.
Clear enough to supply this argument.
attendees › anyOf › 0 › items › email
Attendee email address
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
attendees › anyOf › 0 › items › remove
true removes this attendee. null or false adds or keeps them.
Clear enough to supply this argument.
clearAttendees
true removes every attendee from the event, which cancels their invitations. null or false leaves the attendee list unchanged. Cannot be combined with a non-empty attendees list — pass one or the other.
Clear enough to supply this argument.
seriesUpdateMode
Scope of update for recurring events: single (this occurrence only), all (entire series), or future (this and all future occurrences)
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "update_event",
"title": "Update Event",
"description": "Update an existing calendar event. Only provided fields are changed. Timing updates interpret start as display wall-clock times in timing.timeZone, or the user's time zone when timing.timeZone is null; the event's original storage timezone is preserved. For all-day events, timing.duration is a whole number of calendar days (P1D is one day, P2D is two). list_events and get_events return an exclusive all-day end: a one-day event starting 2026-09-15 has end 2026-09-16, matching duration P1D. An all-day timing.start can be a date (2026-09-15) or local midnight (2026-09-15T00:00:00).",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"id",
"accountId",
"calendarId"
],
"properties": {
"id": {
"type": "string",
"minLength": 1,
"description": "ID of the event to update (obtained from list_events)"
},
"title": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
]
},
"timing": {
"anyOf": [
{
"type": "object",
"required": [
"start",
"duration",
"isAllDay"
],
"properties": {
"start": {
"type": "string",
"minLength": 1,
"description": "Updated start as wall-clock local date-time (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — this is the display wall-clock in timing.timeZone (defaults to the user's time zone). The event's storage timezone is preserved server-side. When timing.isAllDay is true, a date (e.g. 2026-09-15) is also accepted, read as 2026-09-15T00:00:00."
},
"duration": {
"type": "string",
"minLength": 1,
"description": "Updated duration in ISO 8601 duration format (e.g. PT1H for 1 hour, P1D for one all-day calendar day). All-day events use whole-day durations: P1D is one day, P2D is two consecutive days. The exclusive end date is start plus duration (a one-day event on 2026-09-15 has end 2026-09-16)."
},
"isAllDay": {
"type": "boolean",
"description": "Whether this is an all-day event. All-day duration is counted in whole calendar days with an exclusive end (P1D covers only the start date; list_events / get_events return end as the next day)."
},
"timeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone for start/end wall-clock times (e.g. Europe/Zurich). For timed events this is the request/user display zone — not the event's original storage zone."
},
{
"type": "null"
}
],
"description": "Display IANA time zone for interpreting start (e.g. Europe/Zurich). Send null to use the user's time zone. Does not change the event's storage timezone."
}
}
},
{
"type": "null"
}
],
"description": "Timing update — provide all fields together or omit entirely."
},
"privacy": {
"anyOf": [
{
"enum": [
"public",
"private",
"secret"
],
"type": "string",
"description": "Event visibility: public (visible to everyone), private (details hidden), or secret (not visible)"
},
{
"type": "null"
}
],
"description": "Event visibility: public, private, or secret. null is no change; reset it with clearFields."
},
"location": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"description": "Event location name. null is no change; clear it with clearFields."
},
"accountId": {
"type": "string",
"minLength": 1,
"description": "Account ID that owns the event (obtained from list_calendars)"
},
"attendees": {
"anyOf": [
{
"type": "array",
"items": {
"type": "object",
"required": [
"email"
],
"properties": {
"email": {
"type": "string",
"minLength": 1,
"description": "Attendee email address"
},
"remove": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "true removes this attendee. null or false adds or keeps them."
}
}
}
},
{
"type": "null"
}
],
"description": "Merge-patch attendees. Listed attendees are merged with the existing ones; attendees that are not listed are preserved. Adding an attendee sends a calendar invitation. Set \"remove\" to true to drop one attendee, or use clearAttendees to drop all of them. null or [] changes nothing. Attendee names cannot be changed — providers resolve display names from their directories."
},
"calendarId": {
"type": "string",
"minLength": 1,
"description": "Calendar ID that contains the event (obtained from list_calendars)"
},
"clearFields": {
"anyOf": [
{
"type": "array",
"items": {
"enum": [
"description",
"location",
"freeBusyStatus",
"privacy"
],
"type": "string"
}
},
{
"type": "null"
}
],
"description": "Fields to clear: description, location, or freeBusyStatus and privacy (to the default). Do not also send a value for a cleared field."
},
"description": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Updated event description in plain text or HTML. null is no change; clear it with clearFields."
},
"clearAttendees": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "true removes every attendee from the event, which cancels their invitations. null or false leaves the attendee list unchanged. Cannot be combined with a non-empty attendees list — pass one or the other."
},
"freeBusyStatus": {
"anyOf": [
{
"enum": [
"free",
"busy"
],
"type": "string",
"description": "Whether the event blocks availability (busy) or not (free)"
},
{
"type": "null"
}
],
"description": "Whether the event blocks availability (busy) or not (free). null is no change; reset it with clearFields."
},
"seriesUpdateMode": {
"anyOf": [
{
"enum": [
"single",
"all",
"future"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "Scope of update for recurring events: single (this occurrence only), all (entire series), or future (this and all future occurrences)"
}
}
},
"annotations": {
"title": "Update Event",
"readOnlyHint": false,
"openWorldHint": true,
"idempotentHint": true,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description98/100
Creates an unsent email draft in Kai — a reply when replyToEmailId is set, otherwise a new message. For replies, provide replyToEmailId. For new emails, provide accountId, to, and subject. Returns { id, openLink, threadId, status, systemInfo }: id is the Kai draft id that update_email_draft, send_email_draft and discard_email_draft take. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there. Give the user the openLink.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
replyToEmailId
Encoded ID of the email to reply to. Omit to compose a new email.
Clear enough to supply this argument.
accountId
Account ID for new emails (required when replyToEmailId is omitted; obtained from list_accounts)
Clear enough to supply this argument.
subject
Email subject line (required for new emails; inferred from original for replies).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
bodyText
Plain-text body of the email.
Explain the meaning of this value, not just its name.
bodyHtml
Not supported: a draft keeps plain text only. Put the body in bodyText.
Resolve the mismatch between this description and its schema constraints.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
bcc › anyOf › 0 › items › name
Recipient display name as shown in email clients
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
clientToolCallId
Only after an emailDraft client tool call came back unavailable: the call id that its result names. With it, the draft is the same one the app makes for that call.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "create_email_draft",
"title": "Create Email Draft",
"description": "Creates an unsent email draft in Kai — a reply when replyToEmailId is set, otherwise a new message. For replies, provide replyToEmailId. For new emails, provide accountId, to, and subject. Returns { id, openLink, threadId, status, systemInfo }: id is the Kai draft id that update_email_draft, send_email_draft and discard_email_draft take. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there. Give the user the openLink.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"cc": {
"anyOf": [
{
"type": "array",
"items": {
"type": "object",
"required": [
"email"
],
"properties": {
"name": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Recipient display name as shown in email clients"
},
"email": {
"type": "string",
"minLength": 1,
"description": "Recipient email address (e.g. [email protected])"
}
}
}
},
{
"type": "null"
}
],
"description": "Carbon-copy recipients. Belongs only in drafts where the user explicitly mentions cc."
},
"to": {
"anyOf": [
{
"type": "array",
"items": {
"type": "object",
"required": [
"email"
],
"properties": {
"name": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Recipient display name as shown in email clients"
},
"email": {
"type": "string",
"minLength": 1,
"description": "Recipient email address (e.g. [email protected])"
}
}
}
},
{
"type": "null"
}
],
"description": "Primary recipients (required for new emails; inferred from original for replies)"
},
"bcc": {
"anyOf": [
{
"type": "array",
"items": {
"type": "object",
"required": [
"email"
],
"properties": {
"name": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Recipient display name as shown in email clients"
},
"email": {
"type": "string",
"minLength": 1,
"description": "Recipient email address (e.g. [email protected])"
}
}
}
},
{
"type": "null"
}
],
"description": "Blind carbon-copy recipients. Belongs only in drafts where the user explicitly mentions bcc."
},
"subject": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Email subject line (required for new emails; inferred from original for replies)."
},
"bodyHtml": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"description": "Not supported: a draft keeps plain text only. Put the body in bodyText."
},
"bodyText": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Plain-text body of the email."
},
"accountId": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Account ID for new emails (required when replyToEmailId is omitted; obtained from list_accounts)"
},
"replyToEmailId": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Encoded ID of the email to reply to. Omit to compose a new email."
},
"clientToolCallId": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Only after an emailDraft client tool call came back unavailable: the call id that its result names. With it, the draft is the same one the app makes for that call."
}
}
},
"annotations": {
"title": "Create Email Draft",
"readOnlyHint": false,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description98/100
Revises one unsent email draft by its Kai id. Only the fields you give change; the other drafts stay as they are. When the draft has userEditedAt set, the user edited it: read it with list_email_drafts first and keep the user's text unless they asked you to change it. A sent or discarded draft cannot be changed. Returns the draft with its openLink. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there. Give the user the openLink.
The description explains the capability, when to use it, and what to expect.
Inputs83/100
Explain the meaning of this value, not just its name.
id
The Kai draft id (from create_email_draft or list_email_drafts)
Clear enough to supply this argument.
subject
New subject line. Omit to keep it.
Clear enough to supply this argument.
bodyText
New plain-text body. It replaces the whole body. Omit to keep it.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description93/100
Lists the user's Kai email drafts, newest first, with their Kai id, recipients, subject, body, userEditedAt (set when the user edited the draft) and openLink. Discarded drafts are never listed.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
status
draft (default): unsent drafts. sent: drafts that went out.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
threadId
Only the drafts that reply in this email thread.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
origin
Only drafts from this source. emailTriage: suggested replies from email triage. user: drafts that the user started in the Kai apps.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
limit
Maximum drafts to return (default 20, at most 50).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description93/100
Retrieve one or more emails by their IDs. Returns an array in input order. Each entry is either an email object (type: 'email') or an error object ({ id, error }) for IDs that could not be fetched.
The description explains the capability, when to use it, and what to expect.
Inputs58/100
2 of 3 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
ids
Array of email IDs to retrieve (obtained from email listing)
Clear enough to supply this argument.
includeAttachments
Include attachment metadata (name, type, size, attachmentId). Metadata only; file contents are never included.
Explain what happens when this optional value is omitted.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_emails",
"title": "Get Emails",
"description": "Retrieve one or more emails by their IDs. Returns an array in input order. Each entry is either an email object (type: 'email') or an error object ({ id, error }) for IDs that could not be fetched.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"ids"
],
"properties": {
"ids": {
"type": "array",
"items": {
"type": "string",
"minLength": 1
},
"minItems": 1,
"description": "Array of email IDs to retrieve (obtained from email listing)"
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
},
"includeAttachments": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "Include attachment metadata (name, type, size, attachmentId). Metadata only; file contents are never included."
}
}
},
"annotations": {
"title": "Get Emails",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_thread_emails",
"title": "List Thread Emails",
"description": "Retrieve all emails in a thread, sorted oldest-first. Provides the full conversation context for replies drafted with create_email_draft.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"threadId"
],
"properties": {
"threadId": {
"type": "string",
"minLength": 1,
"description": "Thread ID to retrieve all emails for (obtained from email listing)"
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
}
}
},
"annotations": {
"title": "List Thread Emails",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description96/100
Send an unsent email draft by its Kai id (from create_email_draft). A draft that was already sent is not sent again: the result then has alreadySent: true. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there.
The description explains the capability, when to use it, and what to expect.
Inputs97/100
Input documentation gives an agent enough context to populate the arguments.
id
The Kai draft id (from create_email_draft or list_email_drafts)
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "send_email_draft",
"title": "Send Email Draft",
"description": "Send an unsent email draft by its Kai id (from create_email_draft). A draft that was already sent is not sent again: the result then has alreadySent: true. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"minLength": 1,
"description": "The Kai draft id (from create_email_draft or list_email_drafts)"
}
}
},
"annotations": {
"title": "Send Email Draft",
"readOnlyHint": false,
"openWorldHint": true,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description85/100
Send an email from Kai ([email protected]) to at most 10 total recipients. Recipients may be any address. Reply-To is the user's principal email. A Kai attribution signature is appended. Requires the user setting permissionSendKaiEmail: user (default; only the user's addresses), any (anyone), or disallow (blocked). Unset is the same as user. bodyText is sent as plain text and bodyHtml as HTML; markdown is not rendered. create_email_draft and send_email_draft send from the user's mailbox instead.
The description explains the capability, when to use it, and what to expect.
Inputs71/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
subject
Email subject line.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
bodyText
Plain-text body. Sent as text/plain with no markdown rendering. At least one of bodyText or bodyHtml is required.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
bodyHtml
HTML body. Sent as text/html with no markdown rendering. At least one of bodyText or bodyHtml is required.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description89/100
Send a text to the user on the messaging channel they connected: iMessage, RCS, or SMS from the Kai line, WhatsApp, or Telegram. The recipient is the user's connected messaging account; there is no recipient argument. On WhatsApp the message is only delivered if the user wrote to Kai in the last 24 hours; otherwise the call fails and the user must message Kai first. Requires the user setting imessageLinkingEnabled: true. Unset is the same as false. content is at most 4000 characters.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
content
Plain-text body sent to the user's linked phone. At most 4000 characters.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "send_kai_text",
"title": "Send Kai Text",
"description": "Send a text to the user on the messaging channel they connected: iMessage, RCS, or SMS from the Kai line, WhatsApp, or Telegram. The recipient is the user's connected messaging account; there is no recipient argument. On WhatsApp the message is only delivered if the user wrote to Kai in the last 24 hours; otherwise the call fails and the user must message Kai first. Requires the user setting imessageLinkingEnabled: true. Unset is the same as false. content is at most 4000 characters.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"content"
],
"properties": {
"content": {
"type": "string",
"maxLength": 4000,
"minLength": 1,
"description": "Plain-text body sent to the user's linked phone. At most 4000 characters."
}
}
},
"annotations": {
"title": "Send Kai Text",
"readOnlyHint": false,
"openWorldHint": true,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description96/100
Forward an existing email to new recipients. The integration quotes the original below any intro in bodyText or bodyHtml, and sends immediately. create_email_draft stores an unsent draft instead of sending. Provide emailId from search_emails or get_emails. At least one recipient in to, cc, or bcc is required.
The description explains the capability, when to use it, and what to expect.
Inputs81/100
Clarify units, where to obtain this value, or conventions not already expressed in the schema.
emailId
Encoded ID of the email to forward.
Clarify units, where to obtain this value, or conventions not already expressed in the schema.
subject
Optional subject line (defaults to Fwd: plus the original subject).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
bodyText
Optional plain-text intro prepended above the quoted original.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
bodyHtml
Optional HTML intro prepended above the quoted original.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description91/100
Discard (permanently delete) an email draft by its Kai id. A sent draft cannot be discarded. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there.
The description explains the capability, when to use it, and what to expect.
Inputs97/100
Input documentation gives an agent enough context to populate the arguments.
id
The Kai draft id (from create_email_draft or list_email_drafts)
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "discard_email_draft",
"title": "Discard Email Draft",
"description": "Discard (permanently delete) an email draft by its Kai id. A sent draft cannot be discarded. The draft is saved in Kai. The user can review, edit and send it only in the Kai apps. It is not in the user's Gmail or Outlook, and Kai cannot put it there.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"minLength": 1,
"description": "The Kai draft id (from create_email_draft or list_email_drafts)"
}
}
},
"annotations": {
"title": "Discard Email Draft",
"readOnlyHint": false,
"openWorldHint": false,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description79/100
List pending AI notifications for the authenticated user. Filters: source, updatedAfter. Large pages may omit optional fields; those names appear in elidedFields when that happens.
The description explains the capability, when to use it, and what to expect.
Inputs25/100
1 of 3 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
source
Not provided
Missing description. Explain what this value means and how to supply it.
updatedAfter
Only return notifications updated after this local date-time without timezone suffix (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — the value is read in ianaTimeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_ai_notifications",
"title": "List AI Notifications",
"description": "List pending AI notifications for the authenticated user. Filters: source, updatedAfter. Large pages may omit optional fields; those names appear in elidedFields when that happens.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"source": {
"anyOf": [
{
"enum": [
"email",
"meeting",
"workflow"
],
"type": "string"
},
{
"type": "null"
}
]
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
},
"updatedAfter": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "ISO 8601 local date-time without timezone suffix (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — the zone comes from ianaTimeZone, which defaults to the user's configured time zone. To mean a specific zone, send the wall-clock time in that zone and set ianaTimeZone to it."
},
{
"type": "null"
}
],
"description": "Only return notifications updated after this local date-time without timezone suffix (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — the value is read in ianaTimeZone."
}
}
},
"annotations": {
"title": "List AI Notifications",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description81/100
Batch-mutate AI notifications. Each operation names a notification id and an action (resolve, dismiss, snooze, unsnooze, address, patch). Status transitions are FSM-validated. Date-times are wall-clock values read in ianaTimeZone, which defaults to the user's time zone. Returns one success/error result per operation, in order.
The description explains the capability, when to use it, and what to expect.
Inputs77/100
21 of 22 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
operations › items › id
ID of the notification to mutate (obtained from list_ai_notifications)
Clear enough to supply this argument.
operations › items › action
What to do with this notification. "resolve" requires summary; "address" requires summary and addressedBy; "snooze" requires snoozedUntil; "dismiss" and "unsnooze" take no extra fields; "patch" updates content fields. Status transitions are FSM-validated: resolve/dismiss/snooze/address require status=pending, unsnooze requires status=snoozed.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
operations › items › summary
Required for "resolve" (resolution summary) and "address" (how it was addressed). null for every other action.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
operations › items › addressedBy
Required for "address": who addressed the notification. null for every other action.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
operations › items › snoozedUntil
Required for "snooze": local date-time without timezone suffix (e.g. 2025-06-15T09:00:00) when the notification returns to pending. Interpreted in the call's ianaTimeZone, so set that field too whenever the request names a zone. null for every other action.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
operations › items › title
"patch" only: updated title. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › message
"patch" only: updated message. null leaves it unchanged.
Explain the meaning of this value, not just its name.
operations › items › pushBody
"patch" only: plain-text push body. null = no change; clear via clearFields.
Clear enough to supply this argument.
operations › items › importance
"patch" only: updated importance. null leaves it unchanged.
Explain the meaning of this value, not just its name.
operations › items › deliveryMode
"patch" only: updated delivery mode. null leaves it unchanged.
Explain the meaning of this value, not just its name.
operations › items › expiresAt
"patch" only: notification TTL as local date-time without timezone suffix (e.g. 2025-06-15T09:00:00), interpreted in the call's ianaTimeZone. null leaves it unchanged; list "expiresAt" in clearFields to remove it.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
operations › items › relevantUntil
"patch" only: thread time-sensitivity end as local date-time without timezone suffix (e.g. 2025-06-15T09:00:00), interpreted in the call's ianaTimeZone. null leaves it unchanged; list "relevantUntil" in clearFields to remove it.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
operations › items › threadId
"patch" only: linked thread ID. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › eventId
"patch" only: linked event ID. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › meetingSessionId
"patch" only: linked meeting session ID. null leaves it unchanged.
Explain the meaning of this value, not just its name.
operations › items › accountId
"patch" only: linked account ID. null leaves it unchanged.
Explain the meaning of this value, not just its name.
operations › items › emailId
"patch" only: linked email ID. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › workflowId
"patch" only: originating agentic-workflow ID. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › runId
"patch" only: originating agentic-workflow run ID. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › chatMessageId
"patch" only: assistant message ID in the user's primary Kai conversation. null leaves it unchanged.
Clear enough to supply this argument.
operations › items › clearFields
"patch" only: names of fields to clear. null or [] clears nothing. A field cannot be set and cleared in the same operation.
Clear enough to supply this argument.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description95/100
Create an AI notification on Kai's desk, the app's list of items waiting for the user. Default silent; use deliveryMode all-devices to also push. message may be markdown; set plain-text pushBody when pushing. Returns { id }.
The description explains the capability, when to use it, and what to expect.
Inputs42/100
7 of 15 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
title
Plain-text title (no markdown).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
message
Inbox body; markdown OK. Link tasks, events, meetings, and emails with markdown hyperlinks using `openLink` / `webLink` from tool results so Kai's desk renders rich chips. For pushes, prefer pushBody.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
pushBody
Short plain-text push body (no markdown). Used with deliveryMode all-devices; falls back to stripped message if omitted.
Clear enough to supply this argument.
importance
Impact if ignored. Default medium; does not control push.
Clear enough to supply this argument.
deliveryMode
silent (default): inbox only. all-devices: push to registered devices.
Clear enough to supply this argument.
expiresAt
Optional TTL as local date-time without zone (e.g. 2025-06-15T09:00:00). Not yet enforced (MOR-4493).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › threadId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › eventId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › meetingSessionId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › accountId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › emailId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › workflowId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › runId
Not provided
Missing description. Explain what this value means and how to supply it.
relationships › anyOf › 0 › chatMessageId
Assistant message id in the user's primary Kai conversation, set by the chat runner for answers finished while the user was away; the app opens the chat at that message. Chat notifications are created resolved: the push is the delivery, so deliveryMode must be all-devices.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "create_ai_notification",
"title": "Create AI Notification",
"description": "Create an AI notification on Kai's desk, the app's list of items waiting for the user. Default silent; use deliveryMode all-devices to also push. message may be markdown; set plain-text pushBody when pushing. Returns { id }.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"title",
"message"
],
"properties": {
"title": {
"type": "string",
"minLength": 1,
"description": "Plain-text title (no markdown)."
},
"message": {
"type": "string",
"minLength": 1,
"description": "Inbox body; markdown OK. Link tasks, events, meetings, and emails with markdown hyperlinks using `openLink` / `webLink` from tool results so Kai's desk renders rich chips. For pushes, prefer pushBody."
},
"pushBody": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Short plain-text push body (no markdown). Used with deliveryMode all-devices; falls back to stripped message if omitted."
},
"expiresAt": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "ISO 8601 local date-time without timezone suffix (e.g. 2025-06-15T09:00:00). Do not include Z or a timezone offset — the zone comes from ianaTimeZone, which defaults to the user's configured time zone. To mean a specific zone, send the wall-clock time in that zone and set ianaTimeZone to it."
},
{
"type": "null"
}
],
"description": "Optional TTL as local date-time without zone (e.g. 2025-06-15T09:00:00). Not yet enforced (MOR-4493)."
},
"importance": {
"anyOf": [
{
"enum": [
"critical",
"high",
"medium",
"low"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "Impact if ignored. Default medium; does not control push."
},
"deliveryMode": {
"anyOf": [
{
"enum": [
"silent",
"all-devices"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "silent (default): inbox only. all-devices: push to registered devices."
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
},
"relationships": {
"anyOf": [
{
"type": "object",
"properties": {
"runId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"emailId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"eventId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"threadId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"accountId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"workflowId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
},
"chatMessageId": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Assistant message id in the user's primary Kai conversation, set by the chat runner for answers finished while the user was away; the app opens the chat at that message. Chat notifications are created resolved: the push is the delivery, so deliveryMode must be all-devices."
},
"meetingSessionId": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
]
}
}
},
{
"type": "null"
}
],
"description": "Optional links to related entities, the originating agentic-workflow run (workflowId, runId), and the Kai chat answer this notification is about (chatMessageId). Stored as provided; not validated against existing records."
}
}
},
"annotations": {
"title": "Create AI Notification",
"readOnlyHint": false,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description97/100
Search emails across connected accounts, or one account when accountId is set. Use senders, recipients, and participants for exact address matching; use query only for subject or body keywords. To find mail the user sent, filter by recipients with direction outbound; for mail from someone, use senders; for a one-to-one thread, use participants with match only. All filters combine with AND. With no filters, the latest emails are returned. Returns up to limit emails per account (default 10, maximum 25), merged newest first. Check pages for per-account nextPageToken; paginate with that accountId and unchanged filters when more matches are needed. Check status and partialFailures before claiming a search is complete. Results include metadata by default. For full bodies, set includeFullBody with one account and limit <= 10, or use the email lookup tool by ID. hasAttachment flags file attachments; use the email lookup tool for attachment metadata.
The description explains the capability, when to use it, and what to expect.
Inputs87/100
12 of 13 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
query
Optional free-text search for subject/body keywords (provider-specific syntax such as Gmail is:unread is allowed). Not for exact recipient matching — use recipients or participants instead.
Clear enough to supply this argument.
senders
Exact From addresses to match (for example, emails from [email protected]).
Clear enough to supply this argument.
recipients
External To/Cc/Bcc addresses to match. Pair with direction: "outbound" to find emails the user sent to someone.
Clear enough to supply this argument.
participants
External participant addresses from From/To/Cc/Bcc. Use match: "only" for one-to-one conversation history.
Clear enough to supply this argument.
match
How recipients and participants are matched. Defaults to any when address lists are provided.
Clear enough to supply this argument.
direction
Filter by sender direction relative to the connected account. outbound = sent by the user; inbound = received from others.
Clear enough to supply this argument.
after
Only return emails received after this date/time. Accepts YYYY-MM-DD (midnight UTC), an ISO date-time with Z/offset, or a local date-time interpreted in ianaTimeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
before
Only return emails received before this date/time. Accepts YYYY-MM-DD (midnight UTC), an ISO date-time with Z/offset, or a local date-time interpreted in ianaTimeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
limit
Maximum emails per account, at most 25. Defaults to 10, also when below 1. With includeFullBody, use at most 10.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
pageToken
Opaque nextPageToken from pages for one account. Requires accountId; keep the same search filters when paging.
Clear enough to supply this argument.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
accountId
Filter to a specific email account by ID (from connected-account context, search results, or list_accounts). When omitted, searches across all email accounts.
Clear enough to supply this argument.
includeFullBody
When true, include full email body text in results. Requires a single selected account and limit <= 10. Use sparingly; bodies increase token cost.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "search_emails",
"title": "Search Emails",
"description": "Search emails across connected accounts, or one account when accountId is set. Use senders, recipients, and participants for exact address matching; use query only for subject or body keywords. To find mail the user sent, filter by recipients with direction outbound; for mail from someone, use senders; for a one-to-one thread, use participants with match only. All filters combine with AND. With no filters, the latest emails are returned. Returns up to limit emails per account (default 10, maximum 25), merged newest first. Check pages for per-account nextPageToken; paginate with that accountId and unchanged filters when more matches are needed. Check status and partialFailures before claiming a search is complete. Results include metadata by default. For full bodies, set includeFullBody with one account and limit <= 10, or use the email lookup tool by ID. hasAttachment flags file attachments; use the email lookup tool for attachment metadata.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"after": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Only return emails received after this date/time. Accepts YYYY-MM-DD (midnight UTC), an ISO date-time with Z/offset, or a local date-time interpreted in ianaTimeZone."
},
"limit": {
"anyOf": [
{
"type": "integer",
"maximum": 9007199254740991,
"minimum": -9007199254740991
},
{
"type": "null"
}
],
"description": "Maximum emails per account, at most 25. Defaults to 10, also when below 1. With includeFullBody, use at most 10."
},
"match": {
"anyOf": [
{
"enum": [
"any",
"all",
"only"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "How recipients and participants are matched. Defaults to any when address lists are provided."
},
"query": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"description": "Optional free-text search for subject/body keywords (provider-specific syntax such as Gmail is:unread is allowed). Not for exact recipient matching — use recipients or participants instead."
},
"before": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Only return emails received before this date/time. Accepts YYYY-MM-DD (midnight UTC), an ISO date-time with Z/offset, or a local date-time interpreted in ianaTimeZone."
},
"senders": {
"anyOf": [
{
"type": "array",
"items": {
"type": "string"
}
},
{
"type": "null"
}
],
"description": "Exact From addresses to match (for example, emails from [email protected])."
},
"accountId": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Filter to a specific email account by ID (from connected-account context, search results, or list_accounts). When omitted, searches across all email accounts."
},
"direction": {
"anyOf": [
{
"enum": [
"any",
"inbound",
"outbound"
],
"type": "string"
},
{
"type": "null"
}
],
"description": "Filter by sender direction relative to the connected account. outbound = sent by the user; inbound = received from others."
},
"pageToken": {
"anyOf": [
{
"type": "string",
"minLength": 1
},
{
"type": "null"
}
],
"description": "Opaque nextPageToken from pages for one account. Requires accountId; keep the same search filters when paging."
},
"recipients": {
"anyOf": [
{
"type": "array",
"items": {
"type": "string"
}
},
{
"type": "null"
}
],
"description": "External To/Cc/Bcc addresses to match. Pair with direction: \"outbound\" to find emails the user sent to someone."
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
},
"participants": {
"anyOf": [
{
"type": "array",
"items": {
"type": "string"
}
},
{
"type": "null"
}
],
"description": "External participant addresses from From/To/Cc/Bcc. Use match: \"only\" for one-to-one conversation history."
},
"includeFullBody": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "When true, include full email body text in results. Requires a single selected account and limit <= 10. Use sparingly; bodies increase token cost."
}
}
},
"annotations": {
"title": "Search Emails",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description96/100
Retrieve one or more meeting sessions by their IDs. Returns an array in input order. Each entry is either a meeting session object (type: 'meetingSession') or an error object ({ id, error }) for IDs that could not be fetched. Each session carries its summary, the owner's private notes, the brief prepared before the meeting, and the tasks created from its action items with their current status. Set includeTranscript to retrieve full transcript text. Each session's openLink opens the meeting in Kai for its owner only. get_meeting_session_public_link returns a link that other people can open.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
3 of 4 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
ids
Array of meeting session IDs to retrieve (obtained from list_meeting_sessions)
Clear enough to supply this argument.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
includeTranscript
When true, include full meeting transcript text in results. Transcripts can be large (~8.5k tokens per hour of meeting). When false or omitted, transcript is redacted.
Clear enough to supply this argument.
includeTranscriptLineIds
When true, each transcript line starts with its id and the time it was said, as [id time] speaker: "text". Times are in ianaTimeZone, and the date is on the first line and on the first line of each later day. It makes the transcript longer. Without includeTranscript it does nothing.
Clear enough to supply this argument.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_meeting_sessions",
"title": "Get Meeting Sessions",
"description": "Retrieve one or more meeting sessions by their IDs. Returns an array in input order. Each entry is either a meeting session object (type: 'meetingSession') or an error object ({ id, error }) for IDs that could not be fetched. Each session carries its summary, the owner's private notes, the brief prepared before the meeting, and the tasks created from its action items with their current status. Set includeTranscript to retrieve full transcript text. Each session's openLink opens the meeting in Kai for its owner only. get_meeting_session_public_link returns a link that other people can open.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"ids"
],
"properties": {
"ids": {
"type": "array",
"items": {
"type": "string",
"minLength": 1
},
"minItems": 1,
"description": "Array of meeting session IDs to retrieve (obtained from list_meeting_sessions)"
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
},
"includeTranscript": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "When true, include full meeting transcript text in results. Transcripts can be large (~8.5k tokens per hour of meeting). When false or omitted, transcript is redacted."
},
"includeTranscriptLineIds": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"description": "When true, each transcript line starts with its id and the time it was said, as [id time] speaker: \"text\". Times are in ianaTimeZone, and the date is on the first line and on the first line of each later day. It makes the transcript longer. Without includeTranscript it does nothing."
}
}
},
"annotations": {
"title": "Get Meeting Sessions",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description98/100
List meeting sessions for the authenticated user. Filters: after, before (by session start time). Returns sessions with titles, participants, summaries, action items, and status, most recently modified first, or latest start first when after or before is set, capped at 50 per call. If truncated is true the cap was reached and more matched: narrow with after/before before counting or summarizing. Large pages may omit optional fields; those names appear in elidedFields when that happens. Transcripts are redacted — transcript content is available from get_meeting_sessions with includeTranscript: true. Each session's openLink opens the meeting in Kai for its owner only. get_meeting_session_public_link returns a link that other people can open.
The description explains the capability, when to use it, and what to expect.
Inputs50/100
2 of 3 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
after
ISO 8601 local date-time or calendar date; only return sessions that started at or after this time. Do not include Z or a timezone offset — the value is read in ianaTimeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
before
ISO 8601 local date-time or calendar date; only return sessions that started before this time. Do not include Z or a timezone offset — the value is read in ianaTimeZone.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
ianaTimeZone
Not provided
Missing description. Explain what this value means and how to supply it.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_meeting_sessions",
"title": "List Meeting Sessions",
"description": "List meeting sessions for the authenticated user. Filters: after, before (by session start time). Returns sessions with titles, participants, summaries, action items, and status, most recently modified first, or latest start first when after or before is set, capped at 50 per call. If truncated is true the cap was reached and more matched: narrow with after/before before counting or summarizing. Large pages may omit optional fields; those names appear in elidedFields when that happens. Transcripts are redacted — transcript content is available from get_meeting_sessions with includeTranscript: true. Each session's openLink opens the meeting in Kai for its owner only. get_meeting_session_public_link returns a link that other people can open.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"after": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "ISO 8601 local date-time without timezone suffix (e.g. 2025-06-15T09:00:00) or calendar date (e.g. 2025-06-15). Do not include Z or a timezone offset — the zone comes from ianaTimeZone, which defaults to the user's configured time zone."
},
{
"type": "null"
}
],
"description": "ISO 8601 local date-time or calendar date; only return sessions that started at or after this time. Do not include Z or a timezone offset — the value is read in ianaTimeZone."
},
"before": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "ISO 8601 local date-time without timezone suffix (e.g. 2025-06-15T09:00:00) or calendar date (e.g. 2025-06-15). Do not include Z or a timezone offset — the zone comes from ianaTimeZone, which defaults to the user's configured time zone."
},
{
"type": "null"
}
],
"description": "ISO 8601 local date-time or calendar date; only return sessions that started before this time. Do not include Z or a timezone offset — the value is read in ianaTimeZone."
},
"ianaTimeZone": {
"anyOf": [
{
"type": "string",
"minLength": 1,
"description": "IANA time zone (e.g. Europe/Zurich) every local date-time in this call is interpreted in, and that date-times in the response are rendered in. Defaults to the user's configured time zone, so set it whenever the request names a zone — asking for 09:00 UTC means sending 09:00 here with ianaTimeZone \"UTC\", not 09:00 alone."
},
{
"type": "null"
}
]
}
}
},
"annotations": {
"title": "List Meeting Sessions",
"readOnlyHint": true,
"openWorldHint": false,
"destructiveHint": false
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description98/100
Return the public URL of one meeting session, the link to share with people outside Kai. When the meeting has no public URL yet, this call creates one and makes the meeting public. Anyone with the URL can view, without signing in, the title, the participants' names and emails, the summary, the action items, and the full transcript. The page updates when the meeting changes. Private notes are not shown.
The description explains the capability, when to use it, and what to expect.
Inputs98/100
Input documentation gives an agent enough context to populate the arguments.
id
Meeting session ID (obtained from list_meeting_sessions, get_meeting_sessions, or the app). Not a calendar event ID.
Clear enough to supply this argument.
Annotations100/100
All four behavior hints are declared.
{
"title": "Get or Create Meeting Public Link",
"readOnlyHint": false,
"openWorldHint": true,
"idempotentHint": true,
"destructiveHint": true
}
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_meeting_session_public_link",
"title": "Get or Create Meeting Public Link",
"description": "Return the public URL of one meeting session, the link to share with people outside Kai. When the meeting has no public URL yet, this call creates one and makes the meeting public. Anyone with the URL can view, without signing in, the title, the participants' names and emails, the summary, the action items, and the full transcript. The page updates when the meeting changes. Private notes are not shown.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"minLength": 1,
"description": "Meeting session ID (obtained from list_meeting_sessions, get_meeting_sessions, or the app). Not a calendar event ID."
}
}
},
"annotations": {
"title": "Get or Create Meeting Public Link",
"readOnlyHint": false,
"openWorldHint": true,
"idempotentHint": true,
"destructiveHint": true
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description94/100
Search Google Maps for places matching a query. Returns place suggestions with display text (description, suitable as an event location name) and a Google placeId for commute calculations with compute_commute_time.
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
query
Free-text place query (address, venue name, city, etc.).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
limit
Maximum number of suggestions to return (1–10). Defaults to 5.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations75/100
Declare idempotentHint based on the tool’s actual behavior.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description88/100
Compute travel time and distance between two places using Google Directions (driving, walking, transit, or cycling). Supports leave-at / arrive-by timing and returns transit line details when available.
The description explains the capability, when to use it, and what to expect.
Inputs93/100
Remove repeated schema validation from descriptions; focus on meaning and usage context.
origin
Start location: address, place name, latitude,longitude, or a Google place ID written as placeId:<placeId> (the placeId values search_locations returns).
Clear enough to supply this argument.
destination
End location: same formats as origin, including placeId:<placeId> from search_locations.
Clear enough to supply this argument.
mode
Travel mode. Defaults to driving.
Clear enough to supply this argument.
waypoints
Intermediate stops, pipe-separated per Google Directions (addresses or placeId:<placeId> values), or include optimize:true to reorder. Driving/walking/bicycling only.
kai setup, tools, quality score, publishing readiness, and alternatives.
How do I add kai to Claude, ChatGPT, or Cursor?
Add kai to Claude, ChatGPT, or Cursor by connecting its remote MCP URL in each client's MCP settings. In Claude, add a custom connector. In ChatGPT, add an MCP app. In Cursor, add an MCP server under Settings. The MCP URL is https://mcp.hirekai.ai/mcp.
Is kai MCP ready for ChatGPT and Claude?
As of November 19, 2026, kai scores 100/100 on Manufact's MCP publishing checklist. The scan meets ChatGPT publishing checks and meets Claude publishing checks. Passing these checks does not guarantee marketplace approval.
How clear is kai's MCP tool documentation?
As of November 19, 2026, kai scores 78/100 in Manufact's review of tool titles, descriptions, input and output schemas, and behavior annotations. The review covers 41 of 41 discovered tools. The quality score measures documentation clarity; runtime behavior and marketplace readiness require separate checks.
What are the best alternatives to kai?
As of November 20, 2026, alternatives to kai in productivity on the Manufact MCP leaderboard include AdaptlyPost, Notion MCP, Paperpal, Skyward, and VantagePeers.
How can kai improve its MCP documentation?
create_task is the lowest-scoring reviewed tool at 60/100. No output schema is declared. Add one for structured results so clients can interpret returned fields. Open its review for feedback and a copyable fix prompt, then rerun the documentation review after updating your tool metadata.
Not your MCP?
Review any public or OAuth MCP server for protocol compatibility, tool documentation, and ChatGPT and Claude publishing readiness. No account required.