Bank accounts through Plaid. Gmail and Google Drive. QuickBooks and Stripe — around thirty connectors in all. Each one lands data as typed objects in your own isolated tenant, and the documents your inbox already holds — contracts, COIs, invoices, receipts — surface themselves.
The win isn’t the plumbing — it’s what the connections surface on their own. The contract sitting in your inbox, the COI in a shared folder, the receipt on a card feed: each becomes a first-class object in your graph, linked to its party or job, with its facts staged for your confirmation.
From a logo wall to a working graph in four moves — software integrations for a small business that end in confirmed facts, not a pile of raw sync data.
A self-serve center of ~30 connectors, each with granular consent scopes you approve up front. Anything not configured yet says "coming soon" honestly — never a dead button.
Every connection shows live feed health and last-sync. Tokens live in a vault — the database stores only a pointer, never the credential itself.
Extract→Propose reads what the feeds carry — contracts, COIs, invoices, receipts — and turns each into a viewable, linked, searchable object with its extracted facts staged as proposals.
Nothing a connector surfaces enters your working graph until a human approves it. And you can pause, re-consent or disconnect any connection whenever you want.
The connections that turn Command from software into an operating partner — an AI that connects to your bank and email and reads your business the way you would, if you had the time.
Checking, savings, credit cards and loans — balances and transactions land ready to reconcile against your ledger, with deterministic matching. No LLM ever computes a money figure.
Fully available — Command passed Google's CASA security assessment. Connect your inbox and shared folders and the contracts, COIs and invoices already sitting there surface as first-class objects.
Books Mode: QuickBooks stays your source of truth and nothing changes for your accountant. Command connects read-only and becomes the beautiful window — with the intelligence layer on top.
Payments, payouts and disputes flow straight into your receivables — so what customers paid and what the ledger says never drift apart.
Every connection states exactly what it will read before you approve it, and stays visible after: feed health, last sync, and the scopes you granted. Pause a connection, re-consent with different scopes, or disconnectit entirely — any time, from one page.
And the data lands in your own isolated tenant — a dedicated schema under row-level security — while the credentials themselves live in a vault the application database only holds a pointer to.
Gmail — sales@yourco.com · feed healthy · synced 4 min ago.
Money and payments, documents and comms, calendar, web analytics and SEO, social, growth and sourcing, CRM migration — every feed lands in the same ontology, so attribution, the ICP and the agents all read from one place.
Availability is always honest in-product: a connector that isn’t configured for your deployment yet says “coming soon” — never a dead button.
Around thirty connectors across eight families: money and payments (Plaid, Stripe, Square, QuickBooks Online, Ramp), documents and comms (Gmail, Google Drive, Microsoft 365, Slack), calendar (Google Calendar), web analytics and SEO (Google Analytics, Search Console, Microsoft Clarity, Bing), social (LinkedIn, YouTube, TikTok, Instagram, Facebook, X), growth and sourcing (Meta and Google Ads, Google Places, Apollo), CRM migration (HubSpot, Salesforce), and messaging (SendGrid, Twilio). Every feed lands as typed objects in your own isolated tenant.
See it live on your own stack — connect a bank feed, watch documents surface from your inbox, and decide if this is how your tools should have worked all along.