AI in your app
Bool can build AI features into the app you ship — a summarizer, a chat assistant, a “clean up my notes” button, a classifier that tags incoming entries. This is separate from the AI you chat with in the editor: this one runs for your app’s visitors, after you publish.
You don’t need an API key, an account with a model provider, or a billing setup. Ask for the feature in chat — “summarize each review when it’s submitted”, “add a chat helper to this page” — and Bool wires it up.
How it works
The AI call runs on Bool’s servers, not in the browser. Your app sends the prompt, Bool runs it against the model with the platform’s own credential, and the answer comes back. Nothing secret ever ships in your app’s code — which matters, because everything in your app’s source is public to anyone who opens it.
Under the hood this is the bool.ai battery. If you’re writing code yourself,
see bool.ai in the SDK reference
for generate, typed structured output, and stream.
If you want to call another service that you have a key for — a weather API,
a payments provider — that’s bool.fetch plus the Secrets tab, covered in
the same place.
What it costs
AI in a published app spends app credits, the runtime pool that belongs to the workspace that owns the app. Your visitors never pay and never need an account of their own.
A call is charged for what it actually cost to run, so a long prompt or a long answer costs more than a short one. A typical message settles at about 1 app credit. See Plans and billing for each plan’s allotment, how the pool resets, and what carries over.
Two limits apply on every plan, including paid ones:
- Each published app can spend at most a quarter of your pool per day (60 app credits minimum), so one busy app can’t drain the pool and take your other apps down with it.
- A single visitor can make at most 60 AI calls a minute.
When it’s unavailable
If the pool runs out — or the app hits its daily limit — the app keeps working. Only the AI feature stops, and it says so where the answer would have been. You get an email when it happens, so you hear it from us rather than from your users.
Every Bool ships an AiError component for exactly this, and the agent uses it
by default. Your users see “AI features are unavailable” (plus the day it’s
back, when we know it) rather than the billing reason.
Leave the button and the section where they are — an app that hides its AI when the credits run out looks to a visitor like an app that never had the feature.
Availability
AI in published apps has to be enabled for your workspace before it will answer. If you want it turned on, get in touch.