Integration · OpenAI

Pull OpenAI usage and invoices into the unified AICosts.ai dashboard — read-only, under 2 minutes.

0 hr
Default refresh cycle
0 days
Default historical backfill
Read-only
API key permissions required
0
Lines of code changed

Setup in 4 steps

1

Create a read-only API key in the OpenAI console

Settings → API keys → Create new → restrict permissions to read usage and billing. Copy the key.

2

Paste it into AICosts.ai

Dashboard → Platforms → Add OpenAI → paste the key. Backfill defaults to 90 days.

3

Watch it flow in

Usage events and invoice lines land in your dashboard within minutes. Auto-refresh is hourly.

4

Slice by model, environment, or custom tag

Break down GPT-4o vs GPT-4o-mini vs o1 spend. Filter by any usage_metadata tag you passed at request time.

What you get

Separates input tokens, output tokens, and cached tokens

OpenAI's prompt-caching discount (~50% on cached input) is surfaced separately so you can see real cache-hit economics, not a blended number.

Batch-API aware

Batch usage lands in a different bucket and shows its own ~50% rate. You see which workloads are on batch vs real-time.

o1 and reasoning-token tracking

Reasoning tokens on o1-series models are billed separately. AICosts.ai breaks them out so your 'why is o1 so expensive' question has a concrete answer.

Zero application code changes

We never touch your SDK or request path. Read-only billing aggregation, not a proxy.

Use cases

Finding prompt bloat

Week-over-week input-tokens-per-request trend surfaces when few-shots, tool schemas, or retrieved context quietly grew.

Validating model-swap experiments

Tag traffic with experiment=swap_4o_mini, then compare cost-per-request across arms in the dashboard.

Forecasting next month's OpenAI bill

30-day moving average × growth rate beats the 'eyeball the console' method when the number matters for runway.

Frequently Asked Questions

Does AICosts.ai see my OpenAI prompts or completions?+

No. We read from OpenAI's usage and billing APIs, which return aggregate token counts and dollar amounts per model per time period. Prompt content never leaves OpenAI's servers.

Can I track multiple OpenAI organizations or projects?+

Yes. Add each organization key as a separate connection. The dashboard merges them into a single OpenAI view with a per-org filter.

What if I rotate my OpenAI key?+

Paste the new key under the same connection. Historical data is preserved; only the ingestion credential changes.

Does this work for Azure OpenAI?+

Azure OpenAI is a separate integration (via Azure Cost Management) because the billing surface is Azure's, not OpenAI's. We support both side by side.

Connect OpenAI in under 2 minutes

Read-only ingestion. No proxy. No changes to production.

Start tracking your AI spend

Free tier available. Read-only ingestion. No changes to production.