AI spend troubleshooting

Claude cost spikes: causes, checks, and alert policy.

Claude usage spikes when internal tools, product features, or agentic workflows create more model calls than the team expects.

Common causes

What usually moves the Claude bill

  • Internal Claude workflows become part of daily engineering or support operations.

  • Long-context processing repeats across documents or customers without caching.

  • Agent loops and tool calls increase requests per task.

  • Model defaults or routing rules shift toward more expensive usage.

First checks

Triage checklist

  • Compare usage by service, workflow, model, and trace attributes.
  • Review OTEL ingest health and whether events are duplicated.
  • Check recent agent, prompt, and model-routing changes.
  • Separate internal automation from customer-facing product usage.
Alert policy template

Green, amber, red thresholds for Claude

Green

Daily Claude spend is within 10% of baseline and ingest health is current.

Amber

Daily Claude spend is 10-25% above baseline or a workflow starts spending unexpectedly.

Red

Daily Claude spend is more than 25% above baseline or forecast exceeds AI workflow budget.

Next step

Turn this playbook into a daily signal.

StackSpend connects Claude to your cloud and AI cost view with daily Slack or email reporting, anomaly detection, and pace-to-forecast.

Start free
Claude Cost Spikes: Common Causes & Alert Policy — StackSpend