ellivate.ai

Concepts

Pricing & credits

Most of Ellivate is flat-fee — publishing, hosting, KV, blob, schedules, push, sharing, identity. The one thing that draws against credits is LLM inference (`ellivate.reason()`, AI editing, the create-from-prompt flow). Every paid plan includes a generous bucket of managed inference; if you'd rather bring your own provider key, the BYOK plan trades the included bucket for a lower monthly price.

The mental model

Hosting your tool is flat-fee. KV writes, blob uploads, scheduled handlers firing, push notifications, viewers opening the app — all included. You don't meter against per-request usage; you don't worry about a busy week tipping you into a higher bill.

The one place credits enter is when your tool (or Ellivate on your behalf) calls a large language model. That includes:

  • ellivate.reason() — anywhere your tool calls it from a server handler.
  • AI editing — when you describe a change and the EditAgent makes it.
  • Create from prompt — when you describe a new tool on mobile or in the dashboard and Ellivate builds the first version.

Inference has provider costs (Anthropic, OpenAI). Ellivate either pays them and bills you (managed lane) or you bring your own key and pay providers directly (BYOK lane).

Who pays for a tool's AI — the caller

When a tool calls ellivate.reason(), the credits come from whoever triggered the call, not automatically from the person who built the tool. The rule is simple: the AI runs on the caller's credits.

  • A viewer using a shared tool spends their own AI credits. If your friend opens a tool you shared and asks it a question, the inference draws against their bucket — not yours.
  • The @ellivate agent in a Space spends the credits of the member who triggered it. Same for a tool opened from inside a Space: the inference is billed to the member who opened it.
  • You, the owner, are only charged for your own usage — when you use your own tool, or for owner-context calls with no viewer (a scheduled handler firing in the background, a startup task). Nobody else's usage lands on your bill.

This means you can share an AI-powered tool widely without worrying that a popular week runs up your credit. The generous free signup credits cover casual viewer usage, so most people opening a tool you shared never hit a wall.

The plans

PlanPriceManaged inferenceBest for
Free$0A small bucket of trial credits, refreshed every 30 daysKicking the tires. Publishing static-ish tools that don't need an LLM at runtime.
Plus$20 / monthA generous bucket of managed inference, refreshed every 30 daysBuilding tools that actually use ellivate.reason(), doing weekly AI edits, using the create flow.
BYOK$10 / monthYou bring an Anthropic or OpenAI key; you pay providers directly. Ellivate adds zero margin.Power users who already have provider keys, or anyone who wants the cheapest path to high-volume inference.
Premium$50 / monthA much larger managed-inference bucket, 5× higher rate limits, unlimited published tools.Heavy ellivate.reason() usage, busy scheduled handlers, more than 25 tools, or a tool that brushes the standard rate limits.

For exact bucket sizes and the Business tier (team features + higher caps, planned), check the pricing page — that's the source of truth, this doc is the mental model.

Soft caps over hard refusals

The credit model is designed to keep you moving. Two guardrails, neither of which throws a 500 in your tool's face:

  • Warn at 80%. When you've used 80% of your monthly inference credit, the dashboard surfaces a heads-up. Your tool keeps running.
  • Soft over-cap. If you go over the included bucket, calls keep working — the over-cap usage shows up on the next invoice. We'd rather you run a small overage than have your tool break for the people using it.

That said: per-day rate limits exist (60 calls / minute, 1000 / day, per user) to catch accidental loops. If a scheduled handler has a runaway while True: reason(...), you'll hit the rate limit before you hit the credit cap.

BYOK: how it works

Add an Anthropic API key (or OpenAI, when supported) on the dashboard at Account → AI keys. The key is stored encrypted at rest; it's only ever read at the moment a call goes out. Your tool's ellivate.reason() calls now route through your key — Anthropic bills you, Ellivate bills nothing for the inference itself.

Why a $10/month subscription on top? Two reasons:

  • BYOK still uses Ellivate's infrastructure — the agent loops, the caching, the trace persistence, the managed retries when a provider blips.
  • We don't want a free-tier escape hatch where a heavy user pays no contribution to platform costs.

BYOK is a lower-cost lane for someone who's already paying Anthropic; it isn't free.

What about iOS?

Apple's App Store rules don't allow apps to accept subscription payments outside Apple's in-app purchase system. Ellivate's mobile app respects that: subscriptions get managed at ellivate.ai/pricing in the browser, not inside the app. The mobile app surfaces a one-line pointer there. Android keeps full in-app checkout.

What about apps with no LLM calls?

Plenty of useful tools don't reach for inference at all — a household chore tracker, a friend-group photo album, a shared shopping list. Free tier covers those completely. Publish, share, run. The trial credits also cover a few AI edits if you ever want to tweak the tool with English instead of opening the IDE.

Where to manage things

  • Plan + payment — ellivate.ai/pricing (or the dashboard Account menu on web). Stripe Checkout for upgrades; Stripe Customer Portal for downgrade / cancel / payment-method change.
  • AI keys (BYOK) — dashboard Account menu → AI keys. Add Anthropic or OpenAI keys, see which one each tool is using.
  • Spend so far — dashboard Billing tab shows current-period managed inference, the warn/cap indicator, and a per-tool breakdown so you can see where it's going.

What's next