IRISIRIS
The basics

AI spend

Full visibility into your AI API costs across OpenAI and Anthropic, sitting right alongside your AWS spend.

Go to AI Spend in the sidebar to access this view.

Setup

Before any data appears, you need to connect an AI provider admin key in Settings → AI Providers.

  • OpenAI — connect an OpenAI admin key (starts with sk-) to pull usage data from the OpenAI API.
  • Anthropic — connect an Anthropic admin key (starts with sk-ant-admin-) to pull usage data from the Anthropic API.

Once connected, IRIS pulls your usage data automatically. Both providers can be connected simultaneously and are shown as separate tabs.

Anthropic admin keys don't have a read-only scope yet

Anthropic does not currently offer a read-only admin key. The connected key has full admin access — IRIS only ever reads usage and cost data and never modifies your Anthropic account. We'll add scoped support the moment Anthropic ships it.

OpenAI

Key metrics

  • MTD Spend — total OpenAI spend from the 1st of the current month to today.
  • 30d Total — total spend over the last 30 days, with total request count.
  • Month Forecast — projected end-of-month OpenAI spend based on current run rate.
  • Models Used — number of distinct models used, with the top model shown.

Daily spend chart

A bar chart of daily OpenAI spend over the last 30 days overlaid with a trend line.

By model

A sortable breakdown of spend by model. Columns include requests, input tokens, output tokens, cached tokens, and cost. Click any column header to sort.

Cached tokens = direct savings

The cached tokens column shows how much of your input was served from prompt cache, which reduces cost. If it's low and your prompts are stable, prompt caching is the cheapest optimisation you can make.

Other breakdowns

  • By project — spend broken down by OpenAI project.
  • By user — spend broken down by individual user (requires the connected key to have access to per-user pricing data).
  • By API key — spend broken down by individual API key (requires per-key pricing data access).

Organisation

An expandable section showing your OpenAI organisation details:

  • Members — all users in the org with their role and join date.
  • Projects — all projects with member count, API key count, and rate limits per model.
  • API Keys — all admin and project API keys with creation date and last used date.
  • Pending Invites — outstanding invitations to join the org.

Anthropic

Key metrics

  • 30d Total — total Anthropic spend over the last 30 days.
  • Month Forecast — projected end-of-month spend.
  • Models Used — distinct models used, with top model shown.

Daily spend chart

Daily Anthropic spend over the last 30 days.

By model

Sortable breakdown by model including requests, input tokens, cached input tokens, cache write tokens, output tokens, and cost.

Other breakdowns

  • By workspace — spend broken down by Anthropic workspace.
  • By API key — spend broken down by individual API key.

AI spend budgets

To set alerts on your AI spending, create a budget with scope set to AI Spend in the Budgets page. You can set a monthly and/or daily threshold and choose OpenAI or Anthropic as the provider. See Budgets & alerts for full details.