How to
AI Usage Monitoring
Connect your OpenAI and Anthropic accounts to TurboFinOps to track AI API spend, token consumption, and model-level cost breakdown across your organization.
Back to How To GuidesWhat It Does
AI Usage Monitoring lets you connect your OpenAI and Anthropic organization accounts to TurboFinOps and pull billing telemetry -- giving your FinOps and platform teams visibility into AI API spend inside the same platform where you manage cloud infrastructure costs.
It answers questions like:
- Why did our OpenAI invoice jump this month?
- Which models are driving the most cost?
- Which teams or applications are consuming the most tokens?
- Is there an unusual spike in API calls in the last 7 days?
- What is our total AI spend vs. last period?
Available on
Connecting a Provider
OpenAI
- 1In the OpenAI platform (platform.openai.com), go to API keys and create a new key with Usage read permissions.
- 2In TurboFinOps, go to Dashboard - Connections - OpenAI.
- 3Click "Add Usage Connection", enter a display name and paste the API key.
- 4Save. TurboFinOps immediately validates the key and pulls the last 7 days of usage.
- 5Click "Sync Now" to fetch the latest data at any time.
Anthropic
- 1In the Anthropic Console (console.anthropic.com), go to API keys and create a key for monitoring.
- 2In TurboFinOps, go to Dashboard - Connections - Anthropic.
- 3Click "Add Usage Connection", enter a display name and paste the API key.
- 4Save and sync.
Security note
API keys are encrypted at rest using the same mechanism as cloud credentials. They are never logged, never returned in API responses, and used only for pulling usage telemetry from the respective provider.
Reading the Dashboard
The AI Usage monitoring panel shows (per connected provider):
| Metric | Description |
|---|---|
| 30-day cost | Total API spend for the last 30 days in USD |
| Total requests | Number of API calls made in the window |
| Total tokens | Combined prompt + completion tokens consumed |
| Telemetry status | Last sync time and connection health |
| Top models by cost | Ranked list of models with cost and request share |
| Detected signals | Anomalies -- cost spikes, unusual request patterns, high-cost models |
| Usage trend | Day-by-day cost chart for the selected period |
You can add multiple usage connections per provider (e.g. one key per team or project) and view them aggregated or individually.
Anomaly Detection
TurboFinOps automatically surfaces anomalies in your AI usage data. A signal is raised when:
- Daily cost increases significantly compared to the rolling average.
- A model that was not previously used appears and drives high cost.
- Request volume spikes outside normal business hours.
- Token-per-request ratio rises unexpectedly (often indicates prompt bloat or infinite loop patterns).
Detected signals appear in the "Detected Signals" section of the monitoring panel. For deeper context, open Dashboard - AI Anomaly which correlates AI usage signals with cloud infrastructure events (e.g. a deployment that caused a spike).
API Ingestion
If you have a custom AI gateway or want to push usage events from your own systems, you can ingest AI usage data directly via the TurboFinOps API:
POST /api/v1/ai/usage/events -- single event ingestion
POST /api/v1/ai/usage/events/batch -- bulk (up to 300 events/request)
GET /api/v1/ai/usage/summary?days=30 -- usage totals + anomalies
Each event should include: provider, model, promptTokens, completionTokens, costUsd, and a timestamp. See the API reference for the full schema.
AI Usage Monitoring vs. BYOAI (Finn AI)
| AI Usage Monitoring | BYOAI (Finn AI) | |
|---|---|---|
| Purpose | Track AI API spend across your org | Power AI explanations inside TurboFinOps |
| Configured in | Dashboard - Connections - OpenAI / Anthropic | Dashboard - Integrations - Finn AI |
| Data pulled | Usage telemetry (tokens, cost, requests) | Nothing -- sends prompts, receives completions |
| Plans | Enterprise | Enterprise |
| Key reused for | Billing data ingestion only | Generating explanations and summaries |
These are separate surfaces, but both currently require an Enterprise plan. AI Usage Monitoring is for provider billing telemetry, while BYOAI is for in-product explanations.
TurboFinOps
Start with one cloud scope. Prove savings fast.
Connect AWS, Azure, or GCP and get actionable findings, score trends, and auditable remediation paths in minutes.