Skip to main content

Documentation Index

Fetch the complete documentation index at: https://runcrate.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

Keep cloud spending under control without opening the dashboard. Your AI agent can check balances, break down usage, identify waste, and terminate idle resources.

​
”How much have I spent this week and what’s my balance?”

The agent checks billing:
  1. billing_balance — returns current credit balance
  2. billing_usage — returns spending breakdown for the last 7 days
Current balance: $487.30 Last 7 days spending: $112.70
ResourceHoursCost
inference-prod (H100)168 hrs$84.00
training-v3 (A100)12 hrs$19.20
dev-box (RTX 4090)22 hrs$7.70
Storage (3 volumes)—$1.80
Your H100 inference server accounts for 75% of your spend. It’s been running 24/7. At this rate, your balance covers another ~30 days.

​
”Find anything that’s running but not being used and shut it down.”

The agent cross-references instance status with GPU utilization:
  1. list_instances — gets all running instances
  2. ssh_execute — checks nvidia-smi on each running instance
  3. billing_usage — calculates cost per instance
Checked 4 running instances:
InstanceGPUUtilizationRunningCost so far
inference-prodH10042%7 days$84.00
training-v3A10095%12 hrs$19.20
dev-boxRTX 40900%8 hrs$2.80
test-benchRTX 40900%14 hrs$4.90
dev-box and test-bench have 0% GPU utilization. Together they’re burning $0.70/hr. Want me to terminate both?
After you confirm:
  1. delete_instance — terminates dev-box
  2. delete_instance — terminates test-bench
Terminated dev-box and test-bench. Saving 0.70/hr(0.70/hr (16.80/day).

​
”Give me a monthly projection. Can I afford to run an H100 for the rest of the month?”

The agent calculates:
  1. billing_balance — checks current balance
  2. billing_usage — gets daily burn rate from the last 30 days
  3. list_instances — gets current active instances and their per-hour cost
Balance: 487.30∗∗Currentburnrate:∗∗487.30 **Current burn rate:** 2.50/hr (inference-prod H100) + 1.60/hr(training−v3A100)=∗∗1.60/hr (training-v3 A100) = **4.10/hr** At $4.10/hr, you have ~119 hours (~5 days) of runway. If you shut down the A100 after training completes (~6 more hours), the H100 alone burns $2.50/hr — that gives you ~192 hours (~8 days). To run the H100 for the remaining 12 days of the month, you’d need ~720incredits.You′re∗∗720 in credits. You're **233 short**. Options:
  1. Add $250 in credits from the dashboard
  2. Switch to an A100 for inference ($1.60/hr) — covers the full month

​
Tools used in this workflow

ToolPurpose
billing_balanceCheck current credit balance
billing_usageGet spending breakdown by resource and period
list_instancesIdentify all running resources
ssh_executeCheck GPU utilization to find idle instances
delete_instanceTerminate idle resources to stop spending