WorkMatrix
An executive assistant for your inbox, calendar, and daily coordination.
WorkMatrix is an executive assistant that runs the day-to-day: it triages and drafts email, manages the calendar, schedules meetings, and keeps tasks and approvals moving — across Google Workspace and Microsoft 365, in one place. Anything consequential waits for your approval before it goes out.
For Founders, executives, and teams on Google Workspace, Microsoft 365, or both.
Software you
operate.
WorkMatrix is working software, not a black box. You get a workspace built for the human experience, supervision and approval gates that keep you in control, and AgentChat — talk directly to the matrix to direct, question, and review its work.
What it
does.
Inbox triage & drafting
Reads, sorts, and prioritises mail across Google and Microsoft, and drafts replies in your voice — held for approval before sending.
Calendar & scheduling
Manages one calendar across both providers, books meetings, and resolves conflicts the way an assistant would.
Tasks & follow-ups
Tracks what's owed, who owes it, and when — and chases follow-ups so nothing slips.
Approval gate
Consequential actions — sending, accepting, declining — pass an approval gate before they run.
One workspace, two providers
Connect Google and Microsoft accounts side by side; the assistant works seamlessly across both.
Analytics & administration
Workspace analytics and a full administration console for the team.
Built in
modules.
9 modules planned. The module breakdown is published as the matrix nears release.
24 agents · 78 features
Bring
your model.
Every matrix is LLM-agnostic. Run it on SovrinOS for sovereign inference, on SovrinOS Desktop to run the model locally on your own hardware, or connect any major model — you choose per workspace.
AMatrix's sovereign inference — runs the LLM on dedicated GPUs in Montreal. Your data stays in Canada, on AMatrix infrastructure, and is never sent to a third-party model.
Runs the LLM locally on your own hardware — fully sovereign, on-device or air-gapped. The model and your data never leave your machine.
Connect your Anthropic API key from the integration settings.
Connect your OpenAI API key from the integration settings.
Connect your Google AI API key from the integration settings.
Bring any OpenAI-compatible model endpoint through the integration settings.
When you connect a public model like Claude, ChatGPT, or Gemini, AMatrix anonymizes the content before it is sent — identifying and sensitive details are stripped, so they are never exposed to a third-party model. With SovrinOS, nothing leaves AMatrix infrastructure or your own hardware at all.
SovrinOS and SovrinOS Desktop are the sovereign options — running the model locally or on AMatrix's own GPUs — and are never required. LLM connections are configured per workspace.
Plans for
this matrix.
Hand your inbox and calendar to a supervised AI executive assistant — across Google and Microsoft.
50% off your first 90 days New customers who start before October 31st, 2026 — Start and Team / Growth plans. See full pricing →
Ethics
rubric.
Axes scored
- Data minimization
- OAuth scope hygiene
- Vendor TOS compliance
- Log integrity
Prohibited
- Over-broad OAuth scopes
- PII exfiltration via connector
- Bypassing rate limits
Required gates
- Least-privilege scope review
- OAuth audit log per connector
AMatrix agents are decision-support tools. Outputs may contain errors and require human review before action.
Ready when
you are.
See plans above · Business and Enterprise onboarding handled directly with the AMatrix team.