📊 Claude Token Usage Report2026-04-02 #1629
Description
Overview
Period: 2026-04-02T21:10Z (last 24 hours, since previous report at 17:26Z)
Runs analyzed: 1 of 1 Claude-engine workflow run
Total tokens: 79K across all workflows
Estimated total cost: $0.18
One Claude-engine workflow ran during this period: Secret Digger (Claude) — its first appearance in token tracking reports. No smoke-claude or security-guard runs occurred since the previous report.
Workflow Summary
| Workflow | Runs | Total Tokens | Cost | Cache Rate | I/O Ratio | Top Model |
|---|---|---|---|---|---|---|
| secret-digger-claude | 1 | 79K | $0.18 | 99.7% | 31:1 | sonnet-4.6 |
🔍 Optimization Opportunities
-
ℹ️ secret-digger-claude — High cache write rate (50.6%) — This is expected for a short 3-request conversation (1 haiku + 2 sonnet turns). The first sonnet turn writes 38K cache tokens; the second reads them back at 99.7% hit rate. Pattern is healthy for a 2-turn agent run.
- Recommendation: No action needed. If the workflow grows to more turns, cache reuse ratio (read/write) will improve further.
-
ℹ️ No cache on haiku requests — The initial haiku classification call has zero cache tokens. At 123 input tokens and $0.0005 cost, this is negligible.
- Recommendation: Low priority. Add cache breakpoints only if haiku usage scales to >20 requests/run.
Per-Workflow Details
secret-digger-claude
- Run: §23920074040 —
mainbranch - Requests: 3 total
- Models: claude-haiku-4-5-20251001 (1 req), claude-sonnet-4-6 (2 reqs)
- Tokens: 79K total (127 input, 1,230 output, 38,275 cache_read, 39,376 cache_write)
- Cache hit rate: 99.7%
- Cache write rate: 50.6%
- I/O ratio: 31:1
- Avg latency: 10,547ms/request
- Estimated cost: $0.18
Request breakdown:
| # | Model | Input | Output | Cache Read | Cache Write | Duration |
|---|---|---|---|---|---|---|
| 1 | haiku-4-5 | 123 | 95 | 0 | 0 | 1,261ms |
| 2 | sonnet-4-6 | 3 | 1,082 | 0 | 38,275 | 27,542ms |
| 3 | sonnet-4-6 | 1 | 53 | 38,275 | 1,101 | 2,837ms |
Cost breakdown:
- Sonnet cache writes: $0.16 (88%)
- Sonnet cache reads: $0.01 (7%)
- Sonnet output: $0.02 (11%)
- Haiku (all): $0.0005 (<1%)
Workflows Without Token Data
The following Claude-engine workflows did not run during this period:
- smoke-claude — 0 runs since last report
- security-guard — 0 runs since last report
- security-review — 0 runs since last report
Agentic Maintenance (run §23921352791) ran but had no
agent-artifacts— it does not use--enable-api-proxy.
Historical Trend
Previous report: 📊 Claude Token Usage Report 2026-04-02 (issue #1606, created 2026-04-02T17:26Z)
Baseline from previous report (smoke-claude):
- Per-run cost: ~$0.24 | Per-run tokens: ~203K | Cache hit rate: ~99.5%
secret-digger-claude is a new workflow in token tracking — no prior baseline to compare against. Its first run shows healthy caching behavior and a cost of $0.18.
Comparison:
| Metric | smoke-claude (prev) | secret-digger-claude (new) |
|---|---|---|
| Tokens/run | ~203K | ~79K |
| Cost/run | ~$0.24 | $0.18 |
| Cache hit rate | ~99.5% | 99.7% |
| Requests/run | ~6 | 3 |
secret-digger-claude is significantly more token-efficient per run than smoke-claude, likely because it performs a narrower task in fewer turns.
Previous Report
📊 Claude Token Usage Report 2026-04-02 — Issue #1606
References:
- §23920074040 — Secret Digger (Claude), main branch
- §23912847831 — Previous analyzer run
Generated by Daily Claude Token Usage Analyzer · ◷