AI and SaaS savings
CloudYali tracks the cost of Anthropic, OpenAI, Amazon Bedrock and Cursor next to your clouds, but only one AI savings check reaches Cost Savings today. This page lists what exists for each tool, and how to keep AI spend in check in the meantime.
What each tool has
| Tool | Savings checks | In Cost Savings |
|---|---|---|
| Amazon Bedrock | Idle provisioned throughput: in service for 7 days or more with no invocations, priced at its cost over the last 30 days | Yes, under AWS |
| Cursor | Six seat and usage checks, below | Not yet. CloudYali runs them, but the console does not show the results; cloudyali/rajgad#282 tracks adding them |
| Anthropic | None | — |
| OpenAI | None | — |
Cursor checks
CloudYali runs these after each Cursor sync, every 6 hours, over the last 30 days:
| Check | Detects | Monthly savings |
|---|---|---|
| Inactive seat | A member with no activity and no requests for 30 days | The seat price, $40 |
| Underused seat | A member active on 1 to 3 of the last 30 days | Half the seat price |
| Heavy Max Mode use | Max Mode at $50 or more and 40% or more of a member's on-demand spend | The Max Mode premium on those tokens |
| Premium model overuse | $100 or more of spend, 60% of it on models priced 3 times the team's usual | Half the premium on those tokens |
| Context bloat | 50 or more requests averaging 150,000 prompt tokens, $50 or more | 30% of the prompt-token spend |
| Uncapped high spender | No spend limit and spend at or above the team's 90th percentile | Spend above the team's 75th percentile |
Keep AI spend in check
Until more checks reach Cost Savings:
- In Cost Explorer, group AI spend by service or account to see which model, workspace or key costs most. Cost Explorer shows Amazon Bedrock as its own service.
- Set a monthly limit with alerts in Budgets for your AI tools.
- Cost Anomalies flags a day when spend on an Anthropic model or an OpenAI line item jumps above its usual level.
Related
Was this page helpful?