On this page
- 01What you'll do
- 02Where the numbers come from
- 03Open the Explorer
- 04Questions each grouping answers
- 05Investigation: the bill jumped this week
- 06Investigation: who do we charge back
- 07Investigation: are our policies helping
- 08Make the numbers answerable
- 09Read the Dashboard
- 10Export for finance
- 11A weekly review in ten minutes
01
What you'll do
The Explorer turns every governed request into an answer about cost. In about ten minutes you will slice spend, run three common investigations, read the Dashboard, and export the data.
- Break spend down by team, app, model, or key
- Compare windows from the last hour to month to date
- Investigate a jump in the bill, a chargeback, and a policy
- Read what the Dashboard cards mean
- Export the ledger for finance
02
Where the numbers come from
Every request through the gateway is recorded with who sent it, which model and provider served it, how many tokens it used, what it cost, and whether it was blocked. There is nothing extra to instrument.
| Recorded | Shows up as |
|---|---|
| Team, app, environment | Team, App, and Env groupings |
| The virtual key used | Virtual Key grouping |
| Provider and model | Provider and Model groupings |
| Own key or Cloptima credits | Credential Mode grouping |
| Tokens in and out, and cost | Spend, Tokens, and Requests columns |
| Blocked by a policy | Blocked column |
03
Open the Explorer
Go to AI → Explorer. The page has a chart and a table that follow the same two choices.
- 1
Choose Group by
Pick the dimension you want to see.
- 2
Choose a window
Pick 1h, 6h, 24h, 7d, 30d, MTD, or a custom range.
- 3
Read Cost Trend
The chart shows spend over the window for each group.
- 4
Read Spend Breakdown
The table shows spend, requests, input and output tokens, and blocked requests.
- 1Group by
- ProviderModelAppTeamEnvUsed ToolFeature AreaVirtual KeyCredential Mode
- 2Window
- 1h6h24h7d30dMTDCustom
4Spend Breakdown
| Team | Share of Spend | Spend | Requests | In | Out | Blocked |
|---|---|---|---|---|---|---|
| platform | 52% | $312.40 | 48.2k | 31.0M | 6.2M | 12 |
| support | 33% | $198.15 | 31.7k | 20.4M | 3.9M | 0 |
| growth | 14% | $86.02 | 9.4k | 6.1M | 1.1M | 3 |
04
Questions each grouping answers
Pick the grouping that matches the question.
| Group by | Answers | Then |
|---|---|---|
| Team | Which team spends the most | Check that team's policy budget |
| App | Which product or service drives cost | Set a rate limit or output cap on it |
| Model | Which models carry the bill | Look for a premium model on simple work |
| Provider | How spend splits across providers | Compare against your contracts |
| Env | How much is production and how much is testing | Cap non-production with a lower budget |
| Virtual Key | Which integration is responsible | Rotate or restrict that key |
| Feature Area | Which feature a request belongs to | Track cost per feature |
| Used Tool | Which tools agents call | Review the tool allow list |
| Credential Mode | How much runs on your own keys and how much on Cloptima credits | Decide where to move workloads |
05
Investigation: the bill jumped this week
Work from the widest view to the narrowest. Each step narrows the cause.
1Window 7d, group by Model
Did one model grow?
2Group by App
Which app sends it?
3Group by Virtual Key
Which integration?
4Check the Audit tab
Any recent change?
Most jumps trace to one model, one app, or one key. Once you know which, a budget or a rate limit on that policy keeps it from happening again.
06
Investigation: who do we charge back
Finance wants spend by team and environment.
- 1
Group by Team
Choose MTD as the window.
- 2
Add Env
Group by Env to separate production from testing.
- 3
Look for unlabeled spend
Spend with no team means some keys or requests carry no labels. Put team and app on those keys.
- 4
Export
Use Export CSV for the breakdown.
07
Investigation: are our policies helping
Blocked requests are the gateway doing its job. The count tells you where.
- 1
Sort by Blocked
Click the Blocked column header.
- 2
Open the Audit tab
The Blocked count links to it. The Policy Violations card names the rule.
- 3
Decide
A rule blocking real work needs a looser value. A rule blocking mistakes is working.
08
Make the numbers answerable
Spend is grouped by the labels on each request. Put team, app, and environment on your virtual keys, or send them as headers, and every dimension above fills in.
curl https://api.cloptima.ai/v1/ai/chat/completions \
-H "Authorization: Bearer $CLOPTIMA_VIRTUAL_KEY" \
-H "X-Cloptima-Team: support" \
-H "X-Cloptima-App: support-assistant" \
-H "X-Cloptima-Environment: production" \
-H "Content-Type: application/json" \
-d '{"model": "gpt-4o-mini", "messages": [{"role": "user", "content": "Say hello"}]}'You can also add a feature and a workflow label for finer slices. Labels on a key are the most reliable, because every call from that key carries them.
09
Read the Dashboard
The Dashboard summarizes the month and lists what needs attention. Two cards matter most to finance.
| Card line | Meaning |
|---|---|
| Retail cost | What this month's usage costs at list price |
| Contracted cost | What it costs at your approved contract rates |
| Effective blended cost | What you effectively pay after contract rates |
| Credits applied | Credit pools that reduce the blended cost |
| Invoice adjustment | Adjustments from an imported provider invoice |
| Reconciliation drift | How far Cloptima's cost is from the provider invoice |
| Realized Caching Savings | Meaning |
|---|---|
| Provider prompt cache | Savings from your provider's own prompt caching |
| Cloptima cache | Savings from responses served from Cloptima's cache |
| Total saved this month | The two together, with cache read and write tokens |
- Spend spikes
- Budget thresholds
- Contract rates that are about to expire
- Models that have no known price yet, so nothing is silently counted as zero
- Policy blocks
10
Export for finance
Two exports are on the Spend Breakdown card.
| Export | What you get |
|---|---|
| Export CSV | The breakdown you are looking at, in the grouping and window you chose |
| Export FOCUS CSV | The cost and usage ledger in the FinOps Open Cost and Usage Specification, ready for finance tools |
11
A weekly review in ten minutes
A short routine keeps surprises small.
- 1
Group by Team over 7d
Note any team whose spend moved.
- 2
Group by Model
Look for a premium model on traffic that does not need it.
- 3
Sort by Blocked
See which rules are stopping requests, and open the Audit tab for the biggest.
- 4
Check Dashboard actions
Clear spikes, thresholds, and expiring contract rates.