Total calls
Model requests accepted by the gateway in the period, streaming included. Counted per calendar day and reset at midnight.
AI usage and spend, on one screen
Bring company-wide LLM calls, token usage, department rankings, latency, knowledge base hit rate and budget burn together on one big screen. Leadership watches trends, ops watches health, finance watches spend, and everyone works from the same numbers.
Built into TokenHub and ready as soon as you deploy · All figures in the demo are sample data
Free for personal and non-commercial use. Commercial use requires a commercial license.
- 12Core metrics
- 1920×1080Design canvas, scales to fit
- 3–5secAuto-refresh interval
- FkeyOne key to go full screen
Live dashboard preview
METRICS
12 core metrics, clearly defined
Every number comes with a clear definition. All metrics are computed from the gateway's own request logs and drill down to model, department, project and user.
Token usage
Input plus output tokens, as counted by each provider's billing. Local models are estimated the same way so you can compare like for like.
Active users
Unique people who made at least one call in the period, de-duplicated by account or API key owner.
Availability
Successful responses ÷ total requests (excluding client cancellations), sampled every minute and shown as a rolling daily value.
Call trend
Calls per 10 minutes over the last 24 hours, with separate lines for cloud and local models so peaks and offloading are easy to spot.
Model mix
Today's calls by provider and model, shown as both share and count. Local models are listed separately.
Department ranking
Today's token usage rolled up by org structure and ranked, attributed by API key owner and team membership.
Latency & errors
Average, P95 and time-to-first-token latency plus error rate. Crossing a threshold shows an icon and label (Elevated / Critical).
Knowledge base hit rate
Share of RAG queries that return at least one chunk above the relevance threshold, plus query count and top questions.
Cost & budget
Model spend at each provider's list price: a 14-day trend, month-to-date total and budget used, compared with how far into the month you are.
Live request feed
Time, user / department, model, tokens, latency and status for the latest requests. Names and other sensitive details are masked.
Alerts
Latency spikes, error surges, quotas running out, strained nodes, automatic failovers and more, ranked by severity.
WHO IT'S FOR
One screen, three audiences
No more month-end spreadsheet exports or reconciling numbers across teams. The dashboard pulls data scattered across provider consoles into one place, defined consistently and visible in real time.
Leadership sees the big picture
Is AI actually being adopted? Who's using it, and how much? See it at a glance.
- Day-over-day trends for calls, tokens and active users
- Department token rankings that surface your leading AI teams
- Cloud vs. local split to measure how much local models are taking on
Ops keeps things healthy
Latency, error rates and channel status in real time, so you catch problems before users complain.
- Live average, P95 and first-token latency plus error rate
- Failover, rate-limit and node resource alerts as they happen
- Live request feed to pinpoint the model and caller behind a bad request
Finance keeps spend in check
Every dollar of model spend maps to a department and project, and budget pacing is always visible.
- 14-day spend trend plus today's running total
- Budget used vs. month elapsed, with early warning before you overspend
- Breakdowns by provider, department and project for internal chargebacks
* Figures in these cards are sample data.
DISPLAY FIT
One layout, from laptop to ops center
The dashboard is designed at 1920×1080 and scales as a whole, so it keeps the same layout on any screen size or aspect ratio.
-
Proportional 1920×1080 scaling
It scales instead of reflowing, so text, charts and spacing keep their proportions. Nothing shifts or stretches.
-
Any screen, auto letterboxing
Works on 4K TVs, LED video walls, ultrawide monitors and projectors. Non-16:9 screens are centered with bars on the sides or top and bottom.
-
Dark big-screen theme
A dark base with green accents and low-brightness outlines that stays easy on the eyes all day, suited to meeting-room and ops-center lighting.
-
Press F for full screen
Press F or click the button in the top-right corner to enter or exit full screen. No client app to install.
-
Auto-refresh and scrolling
Data refreshes every 3–5 seconds and the feed and alerts scroll on their own. With "reduce motion" turned on, numbers update without animation.
CUSTOMIZATION
Tailor the dashboard to your org
Branding and refresh rate are configurable today. Metric selection and layout templates are in development.
Available
Logo & title
Swap in your own title, logo and subtitle to match your brand guidelines, ideal for lobbies and showrooms.
Available
Refresh rate
Set the refresh interval anywhere from 3 seconds to 5 minutes to balance freshness against system load.
Coming soon
Choose your metrics
Pick the metrics and dimensions to show, for example only one business unit or project.
Coming soon
Layout templates
Preset layouts for executive overview, ops monitoring and cost analysis that you can switch between in one click.
Release timing for upcoming features is announced in the changelog.
FAQ
Live Dashboard FAQ
Is the data really real time?
Calls, tokens, latency and the request feed come straight from the gateway's logs and refresh every 3–5 seconds by default. Cost figures are calculated live from provider list prices; your provider invoices are the final word.
Are the numbers in the demo from real customers?
No. Every number, user and department in the demo is randomly generated in your browser for illustration only and does not reflect any real customer's usage.
Do I need to install it separately? Does it work air-gapped?
No separate install needed. The Live Dashboard is built into TokenHub and is ready as soon as you deploy TokenHub on your servers. It runs fully on single-server, cluster and air-gapped deployments with no external dependencies. TokenHub is free for personal and non-commercial use; for business use, please get a commercial license first.
Can I put it on a meeting-room TV or LED wall?
Yes. The dashboard scales proportionally from 1920×1080, so it looks right on 4K TVs, LED video walls and projectors; non-16:9 screens are letterboxed automatically. Just open the dashboard URL in a browser and press F.
Who can see it? Can department heads see only their own data?
Access is governed by TokenHub's permission system, so you can grant viewing rights by role. Filtering by department will ship with the "Choose your metrics" feature. See Access & Security for details.
Will it slow down if it runs all day?
No. Lists and charts keep a fixed amount of data and drop the oldest entries automatically. Refreshing pauses when the tab is in the background and catches up when it returns, so it's safe to leave running 24/7.
More questions? See the download & install FAQ or contact us.
See where every AI dollar goes
Download TokenHub for free and put company-wide model usage, spend and service health on the big screen as soon as it's deployed. For business use, get in touch about a commercial license.