Skip to main content
Every Hiveku project has a Usage page showing real-time costs broken down by resource — AI tokens, Lambda invocations, S3 storage, CloudFront bandwidth, and more. Good for catching runaway costs early and knowing what to optimize. 
Project > Usage

​
The Three Tabs

  • Overview — cost breakdown by resource this billing period
  • AI — AI-specific usage with per-model and per-mode breakdown
  • Infrastructure — time-series trends for Lambda, bandwidth, cost

​
Overview Tab

Lists each underlying AWS or platform resource your project is using:
ResourceWhat’s measured
LambdaFunction invocations + duration + cost
S3Storage size + request count + bandwidth + cost
CloudFrontCDN requests + data transfer + cost
ACMSSL certificate count + cost
Secrets ManagerSecret count + cost
DatabaseActive connections + storage used
A Total row sums project cost for the current billing period.

​
AI Tab

Where most surprise costs come from — LLM tokens add up fast. The AI tab shows:
  • Total AI cost and total tokens used this period
  • Breakdown by model — Hiveku Mini, Max, Ultra (or the underlying provider names if you’re using BYOK)
  • Breakdown by mode — Code, Designer, Debug, etc. — so you can see which workflows cost the most
  • Recent sessions — a table with session ID, user, cost, tokens, and timestamp
If your AI costs spike, check the Recent Sessions table. It shows the exact prompt and user that drove usage. Often one runaway agent loop or a user pasting huge context accounts for most of a spike. Switch that workflow to Hiveku Mini for simple tasks.

​
Infrastructure Tab

Time-series view for trends, not just current totals:
  • 7-day or 30-day toggle at the top
  • Lambda duration trend — is your code getting slower?
  • Bandwidth trend — are you getting more traffic or bigger assets?
  • Cost trend over time — day-over-day and period-over-period
Helpful for spotting gradual creep — a page that’s 2x the size now as it was a month ago, or a function whose execution time drifted up.

​
Cached vs. Fresh Data

The Usage page caches data for a short window to keep the dashboard fast. Click Refresh in the top right to force a fresh fetch from AWS and the billing service.
Per-project usage here doesn’t replace account-level billing. For invoices, payment methods, and plan management, go to Settings > Billing in the top nav.

​
Using Usage to Optimize

A few practical things you can do after looking at Usage:
1

Identify the biggest line item

Open Overview. Sort by cost descending. Your biggest line is where optimization pays off most.
2

If AI dominates, check modes

Open AI tab. If Code mode on Hiveku Ultra is dominating, see if Hiveku Mini handles the simpler tasks well enough.
3

If bandwidth dominates, compress assets

Check asset sizes in Content > Assets. Images over 500KB are usually over-sized for web. Hiveku auto-optimizes images but custom uploads can slip through.
4

If Lambda duration creeps up, profile

Infrastructure tab’s Lambda trend shows duration. If it’s rising, a slow dependency or growing data set is the cause.

​
Verifying the Numbers

1

Compare to the invoice

Go to Settings > Billing (account-level). Compare this month’s usage totals to what’s on your invoice.
2

Account for timing

Usage in the Usage page is real-time. Invoices bill on a fixed period (monthly, typically). Numbers may not match exactly mid-period.
3

Check for AI BYOK

If you’ve brought your own LLM keys, some AI usage is charged directly to your provider (Anthropic, OpenAI) and not to Hiveku. See BYOK AI.

​
Troubleshooting

Click Refresh in the top right. The cache TTL is a few minutes — refresh forces a live fetch. If it’s still not updating, the upstream billing service may be lagging; try again in 15 minutes.
Usage is real-time; invoices are billed at period close. Timing differences are normal mid-month. At period close, numbers should reconcile. If they don’t, open a support ticket — include screenshots of both numbers.
The project hasn’t deployed yet (or hasn’t had traffic). Infrastructure metrics only start filling in after your first deploy and some real visits.
If you have BYOK enabled, AI calls route to your provider and don’t show in Hiveku’s metering. Your costs appear on your Anthropic/OpenAI dashboard instead. See BYOK AI.
Check the Recent Sessions (AI) or drill into the specific resource. A single runaway agent loop or a misconfigured cron job can account for 90% of a month’s cost. Pause the culprit and contact support if it was caused by a platform issue.

​
What’s Next?

BYOK AI

Bring your own LLM keys to control AI costs separately

Switch Between Accounts

Check usage across your different client accounts