Skip to main content
OpenAI Platform spend sits on a separate bill from AWS, GCP, and Azure. Connect with a read-only Admin API key so organization costs land in with your other billing datasources. This connector imports OpenAI Platform (API usage) organization costs from the Costs API. Costory stores them as a billing datasource, the same as other SaaS billing sources.
Azure OpenAI (Cognitive Services) is billed through Azure. Use the Azure billing connector for that spend. ChatGPT Enterprise and Codex are not on this connector.
1

Create an Admin API key

In OpenAI Platform > Settings > Admin keys, create an Admin API key (sk-admin-...). You must be an organization admin. A project inference key (sk-... or sk-proj-...) will not work. Copy the key immediately. It is shown once.
2

Grant read-only scopes

Grant read-only scopes only. Do not grant write.
3

Connect in Costory

Go to Integrations, click + Add integration, then choose OpenAI under Billing. The connect URL is /integration/new/openai/billing.
4

Enter connection details

Fill Connection name (default OpenAI) and Admin API key. Start date is optional. Leave it empty to use Costory’s default lookback.
5

Validate, then save

Click Validate. Costory calls GET /v1/organization/costs with Authorization: Bearer. Save stays disabled until the Admin key can read organization costs. If this is the only billing source and it is demo data, Costory asks you to confirm before replacing it.
On first sync, Costory backfills up to 12 months of organization costs, depending on what OpenAI retains. New data then follows the usual data refresh cadence. SaaS sources typically appear in about 10 minutes.

​
Field reference

​
What Costory includes

Costory pulls organization costs from OpenAI’s Costs API and maps them into . Typical lines include billed Platform usage for that organization (models, line items, and the project or API key OpenAI returns on the Costs API). How to analyze OpenAI costs
  • Spend by project: one OpenAI project (and project key) per team is the split the Costs API exposes. Map those projects onto a virtual dimension for showback. See Allocate LLM cost per team.
  • Spend by model or line item: group by Service name or SKU
  • Spend by API key or user: names appear when the key has API keys + Projects read. Otherwise you see raw ids.
What Costory does not include
  • Azure OpenAI charges (use the Azure connector)
  • ChatGPT Enterprise or Codex billing
  • Taxes, credits, or invoice adjustments outside the Costs API
  • Per-request application metadata (feature, route, customer). OpenAI’s export does not carry those labels.
Rotate the Admin API key from Integrations when you need a new credential. Keep Usage read on the replacement key.

​
Troubleshooting

Billing Data

All billing providers and shared setup notes.

Anthropic billing

Connect Anthropic Claude Console and claude.ai costs.

Allocate LLM cost per team

Split OpenAI, Anthropic, and cloud LLM spend by team.

Explorer

Filter and group OpenAI spend with the rest of your stack.
Last modified on September 8, 2026