Agent Spend & Savings
16 agent runs · 4 workflows · privacy-safe (no prompts or code stored)
What this audit covers. 16 real Claude Code coding-agent sessions over ~18 days — reading code, editing files, running commands & tests (845 steps; Bash, Edit and Read dominate), across Claude Sonnet, Opus & Haiku.
Total spent
$451.57
across 16 runs
Savings found
$258.94
57% of spend · before validation
Safe to apply now
$114.51
25% · low risk, ready today
Run types
15 / 1
agent runs / monitor checks
Spending by workflow
| Workflow | Type | Runs | Cost | Savings found | Safe now | Savings share | Errors | Biggest opportunity |
|---|
| feature | agent |
4 | $414.65 | $255.88 |
$114.21 |
62% |
12 |
— |
| debug | agent |
2 | $31.16 | $2.59 |
$0.12 |
8% |
0 |
— |
| build | agent |
7 | $4.59 | $0.07 |
$0.06 |
2% |
1 |
— |
| review | agent |
3 | $1.17 | $0.40 |
$0.13 |
34% |
0 |
— |
Where the savings come from
| Compact live turns |
$131.66 |
|
51% of savings |
| Compact live turns (model-verified) |
$97.35 |
|
38% of savings |
| Trim bloated output |
$11.81 |
|
5% of savings |
| Drop unused output |
$5.84 |
|
2% of savings |
| Cheaper model for easy steps |
$5.60 |
|
2% of savings |
| Drop unused data |
$5.35 |
|
2% of savings |
| Skip repeated work |
$1.32 |
|
1% of savings |
| Remove unused tools |
$0.00 |
|
0% of savings |
| Clear old context |
$0.00 |
|
0% of savings |
Same content sent over and over (cache it once)
| Tokens | Seen in runs | What it is | Est. wasted* |
|---|
| 2,266 | 3 | Read output | $0.016 |
| 1,098 | 3 | Read output | $0.008 |
| 631 | 3 | Read output | $0.005 |
| 550 | 3 | tool output | $0.004 |
| 472 | 3 | Read output | $0.003 |
| 327 | 3 | Read output | $0.002 |
*This same text is re-sent across many runs. Store it once instead of paying for it every time.
All runs (click any row for details)
| Run | Type | Cost | Steps | Input load | Cached | Savings found | Useful work | Cost / step |
|---|
feature-03 | agent |
$299.264 | 144 | 223× |
98% | $191.902 |
92% | |
feature-02 | agent |
$65.102 | 113 | 162× |
96% | $50.471 |
93% | |
feature-01 | agent |
$38.571 | 383 | 310× |
98% | $12.614 |
78% | |
debug-02 | agent |
$30.636 | 14 | 84× |
95% | $2.545 |
0% | |
feature-04 | agent |
$11.714 | 73 | 203× |
96% | $0.894 |
80% | |
build-06 | agent |
$2.775 | 1 | 18× |
78% | $0.007 |
100% | |
review-03 | agent |
$0.896 | 29 | 282× |
97% | $0.374 |
16% | |
build-04 | agent |
$0.708 | 2 | 9× |
42% | $0.004 |
100% | |
debug-01 | agent |
$0.523 | 12 | 189× |
93% | $0.043 |
100% | |
build-02 | agent |
$0.453 | 2 | 17× |
66% | $0.006 |
100% | |
build-03 | agent |
$0.402 | 1 | 15× |
0% | $0.000 |
100% | |
review-01 | agent |
$0.167 | 35 | 232× |
95% | $0.012 |
100% | |
build-05 | agent |
$0.156 | 6 | 89× |
94% | $0.001 |
100% | |
review-02 | agent |
$0.108 | 15 | 204× |
92% | $0.013 |
100% | |
build-01 | agent |
$0.092 | 14 | 4274× |
92% | $0.015 |
100% | |
build-07 | monitor |
$0.000 | 1 | 0× |
0% | $0.042 |
0% | |