Meeting assistants
Record → Summarize → Extract
The meeting is the product. What happens afterwards is somebody else's problem, tracked in somebody else's tool.
Currently in pilot
The system of record for team decisions
Turn conversation into coordinated action.
Not a note-taker. Not an agenda timer. Not a hidden answer assistant. MockLive Meetings is the live coordination layer for teams — it holds the purpose of the meeting, surfaces the evidence that settles it, and carries the decision through to verified execution.
We have not finished this product. The web workspace is an early pilot. Desktop capture, SSO, and most integrations are still being built. Ask to join if that is useful anyway.
Pilot on web · macOS and Windows to follow · named-human confirm from day one
MockLive Meetings
Prepare → Coordinate → Decide → Execute → Verify
The meeting is the middle. The product starts before the call and does not close until the decision is done or reversed on the record.
Six systems, one record.
Brief
A meeting is a contract before it is a conversation: objective, decision owner, criteria, pre-reads, definition of done. A meeting missing any of them gets flagged before anyone joins.
Live
A shared visible workspace, not a secret assistant. Agenda, time, decision status, open questions, assumptions, risks, dissent — on screen, for everyone, as it happens.
Context
Permissioned retrieval from the systems you already run, with source, owner, timestamp and confidence attached to every claim. Low-confidence context is labelled, not laundered.
Decisions
A first-class object with an owner, the options considered, the evidence cited, the risk accepted, the dissent recorded, a review date and a rollback trigger. The model proposes. A named human confirms.
Actions
Owner, deadline, dependency, approver, success condition, escalation rule. Written to Linear, Jira, Salesforce or Notion only after a human approves the exact payload.
Memory
Ask why you decided something and get a cited chain of decisions, evidence and reversals — not a model's recollection of a transcript.
Where the category stops.
AreaCategory standardMockLive standard
DecisionsMentioned in the notesFirst-class objects with owner, evidence, dissent and review date
ContextTranscript and basic Q&APermissioned, source-cited retrieval with confidence signals
Action itemsExtracted tasks pushed to a toolHuman-confirmed plans tied to a decision and a success condition
MemorySearch past transcriptsA decision graph with rationale, reversals and outcomes
AI behaviourA silent note takerA visible facilitator with team-approved, admin-allowlisted authority
ReliabilityBest-effort AI summaryConfirmation states, source links, change history, audit trail
Production-ready does not mean never wrong.
No serious system handling speech, collaboration, inference and enterprise data can promise that. The professional standard is different, and we design to it: failures are anticipated, bounded, visible, recoverable and auditable. The pilot is how we find the ones we missed.
The model never confirms
Extraction writes proposals. Only a named human with the right role turns a proposal into a record.
Nothing leaves silently
External writes require the exact payload preview, the permission, and the audit entry. No background pushes.
Consent is versioned
Waiting-room disclosure, per-room policy, guest re-consent. Desktop capture shows the same notice as the browser.
Participation is not a score
Off by default, last-heard timestamps only, never labelled as performance, never exported to HRIS.
Sensitive data is caught first
PII, card and health detection runs before anything reaches the searchable record.
Tenant isolation and residency
Per-org isolation, configurable retention, legal hold, export and delete. Residency is on the roadmap, not in this pilot.
Three apps. One record.
Designed to ship together, not as a web leftover with desktop later. Right now only the browser is in the pilot. Native loopback capture — so no bot has to join your Zoom, Meet, or Teams call — is still a stub.
Any browser
The surface we are piloting now: draft a meeting by talking, share a canvas, confirm as a named human. Guest join and tablet layout are still thin.
Apple silicon & Intel
ScreenCaptureKit loopback, menu-bar tray, signed and notarised, deep links from the calendar. Designed. Not shipping yet.
x64 & arm64
WASAPI loopback, system-tray notifications, MSI for managed fleets. Same record as everywhere else — when it ships.
Your company should not forget why it decided something.
Or write hello@mocklive.ai