AI Usage Monitor
All your AI spend, one widget — 100 % local.
Track your REAL billed AI spend across Anthropic, OpenAI, OpenRouter, Gemini and Mistral against a local estimate. Compact always-on-top widget, detailed dashboard. No telemetry, no backend — keys and data stay on your machine.
No account required · No data sent out · Business logic publicly documented
Your true bill, never blended with the estimate
Each provider has two distinct counters. REAL is fetched from the official billing API (Anthropic Usage API, OpenAI Usage & Costs, etc.). ESTIMATED is computed locally from tokens consumed × embedded price grid. The two are never visually blended: real stays hard, estimated is always labelled "estimate".
Real = what your vendor bills you
Fetched via the official API, with a 60 s local cache to avoid rate limits. No scraping.
Estimated = what your usage would cost on the public rate
Useful when the billing API returns delayed or incomplete data (typically the last 24 hours).
5-hour sliding window
The main chart shows the last 5 hours for immediate feedback, with a one-click access to a 30-day history.
Five providers, one widget
Add an API key — the REAL counter for that provider activates immediately. No key ever leaves your machine.
Anthropic
Claude Sonnet, Opus, Haiku. Usage API + Cost Explorer.
OpenAI
GPT-4.1, GPT-4o, o-series. Usage & Costs API.
OpenRouter
All models available on OpenRouter. /generation endpoint.
Google Gemini
Gemini 1.5 / 2.0 via Vertex API or AI Studio API.
Mistral
Mistral Large, Small, Codestral. La Plateforme API.
Your Claude Code usage, expressed at the public API rate
If you pay for a Claude Code plan, AI Usage Monitor computes what your usage would cost at the public API rate — per model, over the rolling 5-hour window. You get a plain ratio: your plan ÷ value at the API rate. It's a factual reading, not a savings promise.
Example reading
- Claude Code plan
- €100 / month
- Value at the public API rate (this month)
- €312
- Ratio
- ×3.12
Reading: your local usage, priced at the public API rate, would cost ~€312 this month. No conclusion about how much you "save" is displayed — only the raw ratio.
This reading is an equivalent at the public API rate. It does not claim you save — it just shows what your usage would cost on the public grid. No quota percentage is displayed: this mode is reserved for the Pro edition.
100 % local. Zero telemetry.
AI Usage Monitor only talks to the official APIs of the providers you configure — never to a MAG&Cie server or any third party.
Zero backend
All logic runs inside the desktop app. No user account, no server-side authentication.
Zero telemetry
No usage event ever leaves your machine. No analytics, no third-party crash reporter.
Locally encrypted API keys
Keys are stored via the OS-native credential store (DPAPI on Windows, Keychain on macOS).
Local historical data
History and projections live in a local SQLite database, exportable as CSV (Pro).
Free forever, Pro for life
No subscription. One-time purchase, lifetime updates.
Free
€0lifetime
- 1 active provider of your choice
- Always-on-top compact widget
- Detailed dashboard
- Daily budget alerts
- 5-hour sliding window + 7-day history
Pro
Pro€14.99lifetime · 2 seats
- All providers active simultaneously
- Subscription Value (Pro) — equivalent at the public API rate
- 30-day history + monthly projection
- CSV export
- 2 seats per license, lifetime updates
No payment data ever transits through AI Usage Monitor. Payment is handled by Lemon Squeezy (merchant of record, EU VAT included). Your receipt and license arrive by email.
Frequently asked questions
Windows only?
The Windows edition is out. macOS is coming — no committed date, we ship when quality is there. Linux is not on the short-term roadmap.
Do you see my API keys?
No. Keys are encrypted via the OS-native credential store (DPAPI on Windows, Keychain on macOS). None ever leaves your machine — the app doesn't even have a server to receive them.
Difference between REAL and ESTIMATED?
REAL = fetched from the provider's official billing API. It's the accounting truth, potentially delayed by a few hours depending on the provider. ESTIMATED = computed locally from tokens consumed × embedded price grid. Instantaneous but approximate. Both counters are always displayed separately.
Does the "Subscription Value" tell me how much I save?
No — by design. We only show what your local usage would cost AT THE PUBLIC API RATE, in euros and as a ratio (your plan ÷ value at the API rate). No conclusion about "savings" is displayed: the relationship between your plan and the API rate depends on your usage, and it's yours to interpret.
Do you display a quota percentage?
Never. No percentage, no gauge, no "X % of your quota consumed" is shown in AI Usage Monitor — this is a deliberate product decision to avoid false signals.
How does the Pro license work?
One-time €14.99 for life, valid on 2 seats (e.g. your laptop + your desktop). Lifetime updates included. Delivered by email via Lemon Squeezy immediately after payment.
Can I try Pro before buying?
Yes. The Free edition is fully functional with 1 provider — enough to confirm the app delivers on its promise for your setup. Pro unlocks concurrent multi-provider, Subscription Value and CSV export.
Ready to see your true AI bill?
Download for free on Windows, add a provider, open the widget.