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

# Firedog

> AI cost accountability for institutional finance — price every AI call and file it under the workflow, model and desk that spent it.

Firedog turns your vendor AI invoice from one opaque number into a per-call ledger. It prices every LLM call as it happens and files it under the workflow, model and desk that spent it — so finance can attribute the spend, engineering can see why a run costs what it does, and risk can defend the budget with evidence.

## What it answers

<CardGroup cols={2}>
  <Card title="The invoice is one number" icon="file-invoice-dollar">
    The provider bills a single opaque total. You can't tell which desk, workflow or model spent it, so you can't attribute it, defend it, or charge it back.
  </Card>

  <Card title="Agentic runs balloon silently" icon="chart-line">
    Each step re-reads all the context so far, so cost climbs step over step — the re-read tax. Nobody sees why a run costs what it does.
  </Card>

  <Card title="No cheaper-model evidence" icon="scale-balanced">
    You suspect a cheaper model would do the job, but there's no measured cost-vs-quality comparison to prove it.
  </Card>

  <Card title="Nothing to show risk" icon="clipboard-check">
    Finance and risk can't defend the AI budget without a per-call audit trail behind every dollar.
  </Card>
</CardGroup>

## The promise

Price every AI call and file it under the workflow, model and desk that spent it. Firedog reconciles the tagged and untagged spend back to the invoice total, to the cent, and keeps the prompt and response content inside your own VPC — only priced metadata reaches the dashboard.

<Info>
  Firedog runs split-plane. A **collector** inside your VPC prices each call and keeps the full record local; only **metadata** (tokens, cost, latency, attribution) is emitted to the Firedog control plane. See [How it works](/how-it-works) and [Security & data residency](/security).
</Info>

## Start here

<CardGroup cols={2}>
  <Card title="Get started" icon="play" href="/quickstart">
    Route your first call through the collector with the SDK or a base-URL swap, and watch it land on the bill.
  </Card>

  <Card title="How it works" icon="diagram-project" href="/how-it-works">
    The split-plane architecture: the in-VPC data plane, the cloud control plane, and the callId that ties them together.
  </Card>

  <Card title="Security & data residency" icon="shield-halved" href="/security">
    What stays in your VPC, what leaves as metadata, and how sensitivity profiles govern the boundary.
  </Card>
</CardGroup>