AI Spend Tracker
The AI Spend Tracker shows how much money your fleet is spending on AI APIs — broken down by provider, model and server — and lets you set monthly budgets with alerts before you overshoot.
Open it at /ai-compliance/spend or via Compliance → AI Spend Tracker.
What you see
Section titled “What you see”Four KPI tiles across the top:
- This month spend — total USD across every provider so far this calendar month.
- This month calls — total API call count for the same window.
- Budgets — number of active budgets, with an over-threshold indicator.
- 30-day trend — a sparkline of daily spend for the last 30 days.
Below the KPIs, four breakdown tables:
| Table | Grouped by |
|---|---|
| Spend by provider | Provider (OpenAI, Anthropic, Google, Mistral, Groq, …), sorted by cost. |
| Top servers | Individual hostname, top 20 by cost this month. |
| Top models | Provider + model, top 15 by cost this month. |
| Budgets | Every active budget with a progress bar. |
Each row shows both cost (USD) and call count so you can spot cheap-but-chatty workloads as well as expensive one-shots.
What you can do
Section titled “What you can do”- Create a budget — from the inline form on the Budgets card, give it a name, choose a scope (global, per-provider, per-server, or per-service), set a monthly USD limit and an alert threshold percentage.
- Delete a budget — one-click removal from the row’s action.
- Drill into a server — click any hostname to jump to that server’s detail page.
Budgets and alerts
Section titled “Budgets and alerts”Each budget has four fields:
- Name — for your own reference.
- Scope — global (whole tenant), or narrowed to one provider, server or service.
- Monthly limit (USD) — the ceiling to compare against.
- Alert threshold (%) — a percentage of the limit at which the budget flips into a warning state (default 80%).
The progress bar in the Budgets table is colour-coded: green under the threshold, warning-yellow between threshold and 100%, and critical-red once you exceed the limit.
How it works
Section titled “How it works”Spend data comes from daily usage summaries — cost, calls, tokens, provider, model, server — that HostAtlas rolls up from observed AI traffic. Because the summaries are pre-aggregated, the page loads even on large fleets without querying every raw call.
The current month figures use the calendar month in your organisation’s timezone. The 30-day trend rolls a moving window backwards from today so it’s always current.
Budget progress is evaluated on every page load and by a background check that fires alerts when a budget crosses its threshold or limit.
Related
Section titled “Related”- AI Compliance Hub — the parent hub aggregating detections, spend and BOM.
- AI Bill of Materials — the compliance-side inventory of the same AI usage.