Data Model

The MCP server exposes 4 MongoDB collections. Each collection is tenant-isolated — queries are automatically filtered to your account.

Collections overview

Collection	Description	Key use
`signals`	Decisions, actions, accomplishments, key facts extracted from meetings	Find what was decided, who owns what, what happened
`meetings`	Meeting metadata — scheduling, participants, status, recordings	Browse past/upcoming meetings, search by title or attendee
`meetingdata`	Processed transcripts and LLM-generated analysis	Access raw meeting intelligence and AI summaries
`taxonomy`	Organizational entities — people, teams, projects, initiatives	Resolve references, understand organizational context

Relationships

Signals

Extracted insights from meetings. Each signal is a discrete piece of information with a type, content, citation, and optional taxonomy links.

Signal types

A choice or determination made during discussion.

Field	Type	Description
data.title	String	Brief description of the decision
data.rationale	String	Reasoning behind it
data.owner	String	Person responsible
data.impactLevel	String	`high`, `medium`, or `low`

A task or commitment to do something.

Field	Type	Description
data.title	String	Description of the action item
data.assignee	String	Person assigned
data.dueDate	Date	Target completion date
data.status	String	`pending`, `in_progress`, or `completed`
data.priority	String	`high`, `medium`, or `low`

A completed achievement or milestone.

Field	Type	Description
data.title	String	What was accomplished
data.completedBy	String	Person who completed it
data.completedAt	Date	When it was completed
data.impact	String	Description of the impact

Important information or context.

Field	Type	Description
data.title	String	Summary of the fact
data.details	String	Detailed information
data.category	String	Categorization
data.confidence	Number	Confidence score (0.0 to 1.0)

Core fields

Field	Type	Description
_id	ObjectId	Unique identifier
insightType	String	`decision`, `action`, `accomplishment`, or `key_fact`
data	Object	Type-specific content (see tabs above)
createdAt	Date	When the signal was created
updatedAt	Date	Last modification

Citation

Every signal includes a citation linking back to its source:

Field	Type	Description
citation.quote	String	Exact quoted text from the source
citation.surroundingContext	String	Additional context around the quote
citation.speaker	String	Who said it
citation.sourceLocation.startTime	Number	Start timestamp in seconds (audio/video)
citation.sourceLocation.endTime	Number	End timestamp in seconds

Source metadata

Field	Type	Description
sourceChannel	String	Platform: `zoom`, `google_meet`, `slack`, `teams`
sourceId	String	Source document ID (e.g., meeting ID)
timestamp	Date	When the source event occurred
participants	String[]	Participant names from the source

Taxonomy references

Signals can link to organizational entities for context:

Field	Type	Description
people	ObjectId[]	People mentioned or involved
teams	ObjectId[]	Associated teams
initiative	ObjectId	Parent initiative
project	ObjectId	Associated project
task	ObjectId	Specific task

Meetings

Meeting metadata including scheduling, participants, recording status, and processing state.

Core fields

Field	Type	Description
_id	ObjectId	Unique identifier
meeting_id	String	External meeting identifier (from calendar)
meeting_title	String	Display title
meeting_description	String	Description or agenda
meeting_url	String	Join URL
start_time	Date	Scheduled start (ISO 8601)
end_time	Date	Scheduled end (ISO 8601)
status	String	Current status (see below)
summary	String	AI-generated meeting summary
transcript	String	Full text transcript
recording_file	String	URL or path to recording
bot_enabled	Boolean	Whether recording bot is enabled

Status values

Status	Description
`scheduled`	Not yet started
`in-progress`	Currently active
`processing`	Recording/transcript being processed
`completed`	Finished, all processing complete
`failed`	Processing or recording failed
`kicked`	Bot was removed from meeting
`denied`	Bot was denied entry
`cancelled`	Meeting was cancelled

Participants

Each meeting has a participants array:

Field	Type	Description
email	String	Participant email
organizer	Boolean	Whether they organized the meeting
self	Boolean	Whether this is the authenticated user
responseStatus	String	`accepted`, `declined`, `tentative`, or `needsAction`

Meeting Data

Processed transcripts and LLM-generated analysis. A single document can serve multiple meetings.

Field	Type	Description
_id	ObjectId	Unique identifier
meeting_ids	ObjectId[]	Associated meeting IDs
meeting_url	String	Meeting URL
llm_response	Object	AI-generated analysis (flexible structure)
recall_bot_id	String	Recording bot identifier
createdAt	Date	When processed

The llm_response field contains AI-generated analysis and may include meeting summaries, key points, action items, topic extraction, and sentiment analysis. Structure varies by processing configuration.

Taxonomy

Organizational entities that signals reference. Used for context resolution and categorization.

Entity types

Field	Type	Description
email	String	Primary email
role	String	Job title
department	String	Department name
organization	ObjectId	Parent organization

Field	Type	Description
members	ObjectId[]	References to person entities

Field	Type	Description
domain	String	Primary domain (e.g., `acme.com`)
organizationType	String	`internal`, `customer`, `prospect`, or `vendor`

Field	Type	Description
initiative	ObjectId	Parent initiative (for projects)
project	ObjectId	Parent project (for tasks)
status	String	Task status: `pending`, `in_progress`, `completed`
startDate	Date	Start date (tasks)
endDate	Date	Due date (tasks)

Common fields (all types)

Field	Type	Description
_id	ObjectId	Unique identifier
type	String	`person`, `team`, `organization`, `initiative`, `project`, or `task`
name	String	Human-readable name
description	String	Optional description
externalReferences	Object	Links to external systems (Salesforce, HubSpot, Jira, Slack)
createdAt	Date	Created timestamp
updatedAt	Date	Last modified

External references

The externalReferences field links taxonomy entities to external systems:

{
  "salesforce": { "id": "0015000000ABC123", "type": "Account" },
  "hubspot": { "id": "12345678", "portalId": "987654" },
  "jira": { "key": "PROJ-123" },
  "slack": { "channelId": "C0123456789" }
}

Access control

Signals use OR-based access: your user ID in userIds OR your org ID in companyIds. Meetings use AND-based access: both user_id and company_id must match. The MCP server handles this transparently.