> ## Documentation Index
> Fetch the complete documentation index at: https://docs.torchcast.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Torchcast Decision API

> Probabilities, not prose: ask yes/no, multiple-choice and ordinal questions about any text and get calibrated answers back.

The Decision API answers questions about a piece of text with probabilities. You send one `state` (an email, a ticket, a document, a JSON record) and up to 128 named questions about it. Each answer comes back as numbers your code can branch on directly:

* **`noul`**: the probability that the answer to a yes/no question is yes.
* **`choice`**: the most likely of your named options, plus the probability of every option.
* **`score`**: a position on your own ordinal scale, plus the probability of every level.

One endpoint does the work: `POST https://api.torchcast.ai/v1/decisions`. It uses the standard decisions API request and response shape, so moving an existing integration means changing the base URL, the API key and the model id.

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Get a key and make your first request in a few minutes.
  </Card>

  <Card title="Concepts" icon="lightbulb" href="/concepts">
    Questions, answer types, and how to read the probabilities.
  </Card>

  <Card title="API reference" icon="code" href="/api-reference/create-a-decision">
    Every field, with request and response examples to copy.
  </Card>

  <Card title="Migrating" icon="arrow-right-arrow-left" href="/migrating">
    Move an existing decisions integration.
  </Card>
</CardGroup>

## How it works

1. Create an API key in the [console](https://console.torchcast.ai/console). Keys start with `tcd_`.
2. Send a request with your key as a bearer token.
3. Each successful request is charged from your prepaid balance by its billable input tokens. Errors are never charged. See [Billing](/billing).

Prices and per-model limits are published live by [`GET /v1/models`](/models). They are not repeated in these docs.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.