Integrations, AI model & billing
Settings → Account holds three controls: Gmail/Drive connections (one-time Google sign-in, Pro-only), the autofill AI model picker in Preferences, and the Plan row where Stripe billing lives across Free, Lite ($19), Pro ($99), and Scale ($299).
Integrations are personal: you connect your own Google account, and everything riding on it (automations, drafts, the chat assistant's calendar tools) runs under your credentials. Tokens are never shared, so each member powering automations or composing drafts connects their own Gmail. An automation runs under the account of whoever created it (the "powering user"), not the workspace owner.
Three account-level controls live together under Settings → Account, each covered below: connecting Gmail and Drive, choosing your autofill AI model, and managing your plan.
Connecting Gmail and Drive
Both need a Pro plan; toggling on Free or Lite shows "The active workspace requires a Pro subscription to connect automations." Open Settings → Account (the sidebar groups Preferences and Account under "You"). Two toggle rows sit below the Plan row, each subtitled Not connected or the connected email. The consent screen is standard Google OAuth: Gmail asks for mailbox read/modify access plus calendar access; Drive asks for read access to your Drive files.
Gmail powers inbox automations (subject/sender filters that ingest email attachments into cases), composing Gmail drafts when you share a case, and Google Calendar features (the consent screen requests calendar access too, so the chat's calendar tools and Case Data reminder automations share the grant). It does not let the product send email: sharing a case creates a real draft in your mailbox that you Send inside Gmail. Case dates are never pushed to Google Calendar; a case-detail date offers View in calendar (the in-app Calendar) and Download .ics (open in Google, Apple, or Outlook), neither using this connection.
Drive powers Drive import automations (watch a folder, pull new files into intake) plus the automation File storage option (Firebase / Google Drive / Both).
When access is revoked (you revoked the app, credentials rotated, or the grant expired), every automation you power pauses automatically with an amber banner reason: OAuth expired, reconnect Gmail / reconnect Drive; Connected user removed from workspace (the powering teammate was deactivated, so an owner deletes or recreates it); or Paused manually (someone clicked Disable). Recover by re-toggling the integration, completing consent, then re-enabling paused automations from the Automations drawer. A chat calendar tool needing access shows a one-click Connect Google prompt (part of the Gmail consent, not a separate connection). Toggling a row off revokes the token, deletes credentials, and stops any Drive folder watch; cases, files, and data stay untouched, and exporting a case date as .ics still works (it never touches this connection).
Choosing AI models
Model choice is a cost dial, not a feature switch: every model runs every feature. The two selections are deliberately separate so an expensive pick for document extraction never silently makes every chat message pricier. Both persist automatically (no save button); the Preferences row's caption says it plainly: "Used for autofill extraction and field placement. Defaults to Gemini 2.5 Flash Lite."
| Selection | Where | Drives | Default |
|---|---|---|---|
| AI model | Settings → Preferences, AI model row | Autofill extraction, field placement, derived outputs, reflows | Gemini 2.5 Flash Lite |
| Chat model | Chat input footer, next to the Send button | Chat only | Gemini 2.5 Flash Lite |
Chat never inherits the global pick; being high-volume, it defaults cheapest and lets you bump one conversation up. The catalog is intentionally small and cost-first: Gemini 2.5 Flash Lite (0.3x, the only Free-plan model) and Gemini 3.1 Flash Lite (0.5x, newer-generation, stronger spatial grounding for field placement). Internal pinned operations (e.g. chat PDF reading) may use other models; the catalog lists what you pick. On Free, anything else is rejected: "Your plan includes Gemini Flash only. Upgrade to Lite or higher to use other models."
Each call debits your monthly allowance by actual tokens × multiplier, rounded up, so a lower multiplier stretches it further. Autofill runs additionally debit a flat 1,000-token overhead per run.
Chat, autofill, and automated ingestion all meter real usage weighted by the model that ran; metering observes, never blocks mid-flight.
Plans, seats & billing
The boundary that matters sits between Lite and Pro. Free/Lite are template-and-autofill tiers; Pro turns on the dashboard, cases, map + calendar views, additional workspaces (the first is automatic; more, plus editing/deleting workspaces, needs Pro), Gmail/Drive connections, team seats, the client portal, and included batch autofill (an add-on on Lite).
| Free | Lite | Pro | Scale | |
|---|---|---|---|---|
| Price/mo | $0 | $19 | $99 | $299 |
| Templates | 3 | 15 | Unlimited | Unlimited |
| AI tokens/mo | 500K | 2.5M | 8M | 15M |
| OCR pages/mo | 20 | 500 | 5,000 | 15,000 |
| Storage | 500 MB | 5 GB | 50 GB | 500 GB |
| Max upload | 10 MB | 50 MB | 250 MB | 1 GB |
| AI models | Flash Lite only | All | All | All |
| Dashboard/cases | No | No | Yes | Yes |
| Gmail/Drive | No | No | Yes | Yes |
| Batch autofill | No | Add-on | Included | Included |
Annual billing on paid plans saves two months (= 10x monthly). Free/Lite users get one 14-day Pro trial ("Start 14-day free trial").
Seats (Team panel, an expandable row in Settings → Account, "Manage members, seats, and invites"; shows Plan / Limit / Used / Available with pending counts called out): Pro includes 2, Scale 10, Free/Lite single-seat. The owner's seat counts; each active member and each pending invite consumes one (reserved until accepted, declined, or expired); portal clients never do. Extra seats cost $15/mo on Pro ($25 on Scale): set the count, click Update Seats (Stripe-billed, auto-prorated). With none free, inviting fails: "No unassigned seats available. Add seats in Billing to invite more teammates." Team invites are real emails delivered to the invitee's inbox, unlike case sharing (which only composes a Gmail draft for you to send).
Usage panel (expandable row): AI tokens (this month) bar, Pack balance (extra tokens you've purchased), OCR pages (this month), Storage (total). Plans carry a per-autofill token cap and a daily autofill count cap as burst protection. Paid plans over allowance get Buy 1M more tokens, $15, topping up your balance without a plan change.
Plan row controls Stripe: Upgrade opens Checkout; Manage billing opens the customer portal (payment method, invoices, billing cycle, cancellation); Downgrade confirms "Schedule downgrade to Free at the end of your billing period?" plus "You keep access to your current plan until then." A complimentary plan (granted manually, no Stripe subscription) shows "Complimentary access" and no billing buttons. Downgrading never deletes data; Pro surfaces just lock until you upgrade again.
Stripe lifecycle triggers three real sends: the upgrade welcome (plan name, price, what's now unlocked, fired after checkout completes), the subscription receipt (plan, billing cadence, amount, alongside the welcome), and the cancellation confirmation (when the subscription is cancelled or ends). These plus team/collaborator invites are the only mail the product sends; case sharing always composes a draft you send yourself.