Agents get genuinely useful when they can act on your real calendar and contacts — offering appointment slots, recognising callers, booking visits. The connection that makes it possible is made once, at organisation level, on Integrations: no re-authorising per agent, and no credentials anywhere near the AI.
Connect your account
You need to be an organisation owner or admin, on a Pro or Scale plan — Starter shows a See plans card instead.
The sidebar's Integrations page shows one card per provider. Press Connect Google Workspace or Connect Microsoft 365.
polite.ai's own consent modal comes first. Enter the account to connect — it can be a different account from the one you sign in with — and tick the access to grant: Calendar (see and edit events) and Contacts (look up callers), with Mail appearing as an option where it's offered. At least one is required.
Continue to Google/Microsoft sends you to the provider's own consent screen to approve exactly what you ticked. You land back on Integrations with the connection Active.
polite.ai stores a refresh token, so you won't be asked to sign in again — the connection keeps working until you disconnect it or the provider requires re-authorising.
Managing the connection
An Active card shows Connected as your account's email, plus a chip for each granted scope, so it's always clear what access exists. From there:
- Manage access re-opens the consent modal with your current scopes pre-ticked — add or remove access and re-approve with the provider in one pass.
- Reconnect appears as a warning pill when the connection needs re-authorising — one click through the same consent flow restores it.
- Disconnect removes the connection and ends polite.ai's access — the stored tokens are deleted immediately. Agents that depend on it lose those tools until you reconnect.
What your agents can do with it
Once connected, the tools appear in the Builder — calendar lookups, event creation and updates, contact search, and appointment booking through Calendar booking.
Connecting alone gives no agent anything. Each team's tools need your explicit approval: when a team you're building uses Workspace tools, an Enable integration tools step appears in the Builder for an admin to confirm. Read tools are ticked by default; anything that sends or changes things carries a write badge and stays off until you tick it, and enabling mail lets you restrict which domains the agent may ever email.
Teams holding Google or Microsoft tools may only run on approved model providers — OpenAI, Anthropic, Google and Ultravox — so connected-account data never reaches a provider without a no-training commitment. See Models & platforms.
What we ask for, and why
Each scope is requested for one visible purpose: calendar access to offer slots and create bookings, contacts access to recognise callers and optionally save new ones. The AI itself sees only transient, feature-scoped results — offerable slots, a matched caller's name — never your wider account, and never any tokens. The full picture of how connected-account data is handled is in Security, privacy & your data.