Skip to content
Vrge

Managed AI (Optional Hosted Tier)

Managed AI is an optional monthly subscription. We host a cost-optimized inference proxy at ai.getvrge.com, you pay a flat monthly fee, and the app stops worrying about API keys, rate limits, and per-provider pricing. Everything else in Vrge works identically whether you subscribe or bring your own key. Not a single observer feature, extractor, or surface requires Managed AI to function.

If you just subscribed

  1. Find your license key in the receipt email from Lemon Squeezy. It looks like XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX (a UUID).
  2. Open Vrge and go to Settings → AI & Intelligence → Managed AI.
  3. Paste the key and click Activate. The provider dropdown in the AI panel gains a new Vrge Managed option and becomes the default.
  4. On the next observer run, the token meter starts tracking. You can see it in real-time at the top of the AI settings panel.

The three tiers

  • Starter: $10/mo · 1.5M tokens · about 15 new emails a day. One mailbox at a normal pace.
  • Pro: $25/mo · 4M tokens · about 40 new emails a day. One busy mailbox, or two quiet ones.
  • Power: $75/mo · 12M tokens · about 120 new emails a day. Three or four busy mailboxes running the observer all day.

Why email and not team size? Because email is what spends the quota. Of the sources you connect, only your mailboxes are read by the AI — bank transactions, calendar events, files and accounting invoices are extracted structurally and cost nothing. So a solo founder buried in mail can use more than a fifteen-person studio where three people connect an inbox. Asking questions in the command palette and drafting replies draw on the same quota, though far more slowly than the observer does.

These figures cover new mail arriving after you connect, plus the default 7-day first scan. Choosing a longer scan window in Settings → Sources is a one-off charge against the month you run it — a 90-day scan of a busy mailbox can use a whole month's tokens at once. Nothing extra is ever billed: at the cap the observer simply falls back to your own key or to Ollama.

When you hit your monthly cap, Managed AI stops forwarding calls and the app falls back to your BYO key or Ollama for the rest of the month. You will never see an overage charge. Hard quota, not “reasonable usage.”

Upgrading mid-cycle

If your projected usage runs past your budget for the period, Settings → AI & Intelligence → Managed AI shows the projection with an upgrade link to the next tier up (“Upgrade Starter → Pro”). Lemon Squeezy prorates the price difference and credits your quota to the new tier for the remainder of the current billing period. No phone calls, no sales team.

Team admins: watching consumption

In team mode, an admin on any active tier sees a top-consumers table for the current period, with a hard-cap box next to each member. (It is a team-admin view, so a solo user never sees it.) By default every member draws freely from the org quota. There is no automatic per-person split. If one heavy user is a concern, set a hard cap for them and the proxy stops that person before the org quota is spent, so nobody can starve the rest of the team.

Privacy

  • Redaction stays on. The client redacts PII before the call; the proxy re-checks and refuses unredacted payloads when the source is configured redact-required. Belt-and-suspenders by design.
  • No prompt or response logging on the proxy. We record token counts, model name, task type, timestamp, and redaction status. Nothing else. The privacy invariant is enforced at the schema level: there is no column for content in the usage log, so no code path can leak it.
  • Self-hostable. Privacy-max teams (legal, medical, airgapped) do not need the proxy: run a local model, or use your own provider key, and nothing is routed through Vrge.

Cancelling

Cancellation is one click through the Lemon Squeezy customer portal linked on your receipt email. You keep the days you have already paid for: access runs to the end of the current billing period and then simply stops, with no renewal and no refund of the period you are in. No retention nags, no “are you sure” modals, no win-back emails. When the period ends, the app falls back to your BYO key or Ollama automatically. Nothing else changes.