AI Spend Intelligence

Your AI bill is high. One rule is probably causing most of it.

Paste your agent logs. Find out which governance gap is burning your budget. Free. Nothing saved. Nothing leaves your browser.

OpenAI Anthropic Google xAI AWS Bedrock Azure OpenAI

Why is my AI spend so high?

AI costs do not scale with usage. They scale with governance gaps.

Most organisations assume the token bill grows because the business is growing. Sometimes it does. But the bigger driver is almost always a missing or poorly written governance rule causing agents to loop, retry, hallucinate and over-provision.

One bad rule can trigger thousands of wasted steps. The spend looks like a usage problem. It is a governance problem.

Which rule is costing me the most?

That is the exact question nobody could answer. Until now.

Every cost dashboard on the market tells you the total bill. None of them tell you which governance rule caused it. Headlights reads your agent execution traces, detects the waste patterns, names the rule behind each one and puts a dollar figure on it.

Not your agents cost $4,200 this month. The answer is your missing circuit breaker rule cost $2,052. Fix that first.

What does the output look like?

A plain English report you can put in front of your board.

Here is an example. This is exactly what you get when you run your own traces below.

Governance Rule Attribution Example output
Total spend
$2.26
8 agents analysed
Attributable waste
$2.27
100.5% of spend
Governance gaps
9
patterns detected
HIGH
Premature Termination
Incomplete completion criteria
3 triggers
$0.9348
HIGH
Hallucination Recovery
Ambiguous output validation rule
5 triggers
$0.6007
CRITICAL
Cascade Failure
Missing inter-agent boundary rule
2 triggers
$0.2813
CRITICAL
Redundant Retrieval
No caching or retrieval rule
4 triggers
$0.2472
MEDIUM
Over Provisioned Model
No model selection rule
5 triggers
$0.0196
Top recommendation
Fix Premature Termination first. It has triggered 3 times and cost $0.9348. Root cause: incomplete completion criteria.

Is my data safe?

Your logs never leave your browser.

The entire analysis runs in JavaScript on your machine. Nothing is sent to a server. Nothing is stored. Paste your traces, run the analysis, close the tab. Gone.

For critical infrastructure and regulated environments this matters. Your agent logs contain operational reasoning that is not for sharing. This tool is designed so you never have to.

🔒
Zero data transmission. No server calls. No analytics on your trace data. No storage of any kind. The attribution engine runs entirely in your browser using published provider rate cards. Paste, analyse, close the tab.

Run your own traces

How do I use it?

Paste your agent trace JSON below. Or load the example data to see the tool in action across all six providers.

Agent traces (JSON)
Nothing leaves your browser
Attribution report
Severity
Rule and governance gap
Agents affected
Triggers
Cost (USD)