Overview
Period: 2026-05-23 19:48 UTC to 2026-05-24 09:10 UTC
Runs analyzed: 50 (46 had token data)
Total tokens: 27.1M across all workflows
Estimated total cost: $0.00 (cost tracking not yet implemented)
Total Actions minutes: 285.0 min
Total turns: 46
Workflow Summary
| Workflow |
Runs |
Total Tokens |
Avg Tokens |
Cost |
Avg Cost |
Turns |
Errors |
| Build Test Suite |
5 |
9.9M |
2.0M |
$0.00 |
$0.00 |
5 |
0 |
| Test Coverage Improver |
1 |
2.6M |
2.6M |
$0.00 |
$0.00 |
1 |
0 |
| API Surface & Export Audit |
3 |
2.6M |
853K |
$0.00 |
$0.00 |
3 |
0 |
| Test Coverage Reporter |
4 |
2.1M |
520K |
$0.00 |
$0.00 |
3 |
1 |
| Duplicate Code Detector |
1 |
1.7M |
1.7M |
$0.00 |
$0.00 |
1 |
0 |
| CI/CD Pipelines Gap Assessment |
1 |
1.3M |
1.3M |
$0.00 |
$0.00 |
1 |
0 |
| Smoke OTel Tracing |
3 |
1.3M |
434K |
$0.00 |
$0.00 |
3 |
0 |
| Smoke Services |
7 |
1.2M |
175K |
$0.00 |
$0.00 |
7 |
0 |
| Smoke Copilot |
7 |
855K |
122K |
$0.00 |
$0.00 |
7 |
0 |
| Dependency Security Monitor |
1 |
771K |
771K |
$0.00 |
$0.00 |
1 |
0 |
| Smoke Copilot BYOK |
6 |
760K |
127K |
$0.00 |
$0.00 |
6 |
0 |
| Issue Duplication Detector |
3 |
665K |
222K |
$0.00 |
$0.00 |
3 |
0 |
| Daily Claude Token Analyzer |
1 |
387K |
387K |
$0.00 |
$0.00 |
1 |
0 |
| Refactoring Scanner |
1 |
336K |
336K |
$0.00 |
$0.00 |
1 |
0 |
| Smoke Chroot |
3 |
322K |
107K |
$0.00 |
$0.00 |
3 |
0 |
| Firewall Issue Dispatcher |
3 |
226K |
75K |
$0.00 |
$0.00 |
3 |
0 |
🔍 Optimization Opportunities
1. Build Test Suite — 2.0M tokens/run average
Token Analysis: The highest token consumer, using 9.9M tokens across 5 runs (36% of all tokens). This workflow builds and tests the entire project.
Recommendations:
- Consider splitting into smaller focused workflows (unit tests vs integration tests)
- Review if all tests need to run on every trigger, or use conditional execution
- Examine agent prompt length and MCP tool surface area for reduction opportunities
- Pre-compute static analysis results in setup steps to reduce agent investigation time
Trend: Increased from 9.0M (4 runs) in previous report to 9.9M (5 runs) — one additional run but token efficiency similar.
2. Test Coverage Improver — 2.6M tokens in single run
Token Analysis: Single run consumed 2.6M tokens, suggesting extensive analysis or iteration.
Recommendations:
- Review the agent's search strategy for uncovered code paths
- Provide more specific guidance in workflow prompt to reduce exploration
- Consider caching coverage analysis results if this runs frequently
- Break down into smaller incremental improvements rather than comprehensive audits
Trend: Increased from 2.3M in previous report — 13% increase suggests growing complexity.
3. API Surface & Export Audit — 853K tokens/run average
Token Analysis: 2.6M tokens across 3 runs. Likely scans many files for API boundaries.
Recommendations:
- Pre-filter files to only public API surfaces before agent analysis
- Cache module dependency graphs to reduce repeated analysis
- Use static analysis tools before agent review
Trend: Increased from 2.3M (3 runs) to 2.6M (3 runs) — 13% increase per run.
4. CI/CD Pipelines Gap Assessment — 1.3M tokens
Token Analysis: Single comprehensive analysis run, slightly lower than previous report (1.5M).
Recommendations:
- For periodic assessments, analyze only recent changes incrementally
- Pre-generate workflow inventory to reduce discovery time
View Detailed Per-Workflow Analysis
Build Test Suite (5 runs, 9.9M tokens)
Runs:
Pattern: Consistent 2.0-2.4M token usage per run. All complete in 1 turn (efficient, no retries).
Impact: Highest total token consumer. Small optimizations would have significant cost impact.
Test Coverage Workflows
Test Coverage Improver (1 run, 2.6M tokens):
- Single deep analysis of uncovered code paths
- §26334032177 - 2.6M tokens, 1 turn
Test Coverage Reporter (4 runs, 2.1M tokens):
- Average: 520K tokens/run
- More efficient than Improver (reporting vs. improving)
- 1 run had errors (may have affected efficiency)
Smoke Test Workflows
Smoke tests are generally efficient, averaging 100-450K tokens/run:
- Smoke Chroot (3 runs): 107K avg — most efficient
- Smoke Copilot (7 runs): 122K avg — consistent, frequent
- Smoke Copilot BYOK (6 runs): 127K avg — similar to standard
- Smoke Services (7 runs): 175K avg — more services = more tokens
- Smoke OTel Tracing (3 runs): 434K avg — highest (tracing validation complexity)
These workflows are well-optimized for their purpose. No optimization needed.
Analysis & Advisory Workflows
One-time or infrequent workflows for code quality:
- Duplicate Code Detector (1 run): 1.7M tokens
- Dependency Security Monitor (1 run): 771K tokens
- Issue Duplication Detector (3 runs): 665K total, 222K avg
- Refactoring Scanner (1 run): 336K tokens
- Daily Claude Token Analyzer (1 run): 387K tokens
These are complex analysis tasks with appropriate token usage for their scope.
Workflows Without Token Data
The following workflows had runs but reported zero tokens — they may not have api-proxy instrumentation enabled:
- Daily Copilot Token Usage Analyzer (1 run) - This workflow (currently running)
- Daily Claude Token Optimization Advisor (1 run)
Note: Token tracking requires the --enable-api-proxy flag in awf configuration. These workflows may be using direct API calls or running outside the firewall.
Historical Trend
Comparison to Previous Report (#3622):
| Metric |
Previous (2026-05-22 to 05-23) |
Current (2026-05-23 to 05-24) |
Change |
| Total runs |
50 (47 with tokens) |
50 (46 with tokens) |
Same |
| Total tokens |
24.8M |
27.1M |
+9.3% ↗️ |
| Avg tokens/run |
528K |
589K |
+11.6% ↗️ |
| Total turns |
47 |
46 |
-2.1% |
| Action minutes |
243.0 |
285.0 |
+17.3% ↗️ |
Key Observations:
- Token usage increased 9.3% despite similar run count — suggests workflows are doing more work or hitting more complex scenarios
- Action minutes increased 17% — longer runs correlate with higher token usage
- Turns remained stable — efficiency per turn is consistent
- Build Test Suite grew from 9.0M to 9.9M (+10%) — primary driver of increase
Trending Up 🔺:
- Build Test Suite: +10% token usage
- Test Coverage Improver: +13% token usage
- API Surface & Export Audit: +13% token usage
Action Required: Monitor Build Test Suite closely — it's the largest consumer and growing. Consider optimization strategies listed above.
Key Takeaways
- Build Test Suite dominates (36% of all tokens) — optimization here has biggest impact
- Token usage trending up (+9.3% vs yesterday) — monitor for continued growth
- Smoke tests remain efficient (100-450K tokens) — no action needed
- One-time analysis workflows expensive (1-2.6M tokens) but appropriate for scope
- Cost tracking not implemented — all costs show $0.00 (needs api-proxy enhancement)
- High completion rate — 1 error across 50 runs (98% success)
Recommendations
Immediate Actions:
- Review Build Test Suite for splitting or conditional execution
- Monitor token growth trend — investigate if +9% continues
- Enable cost tracking in api-proxy for accurate cost reporting
Long-term Optimizations:
- Cache analysis results for incremental workflows
- Pre-filter files before agent analysis (static first, agent second)
- Set token budgets/alerts for workflows exceeding thresholds
Previous Report
- #3622 — 2026-05-23 report (24.8M tokens across 50 runs)
References:
Generated by Daily Copilot Token Usage Analyzer · ● 4.4M · ◷
Overview
Period: 2026-05-23 19:48 UTC to 2026-05-24 09:10 UTC
Runs analyzed: 50 (46 had token data)
Total tokens: 27.1M across all workflows
Estimated total cost: $0.00 (cost tracking not yet implemented)
Total Actions minutes: 285.0 min
Total turns: 46
Workflow Summary
🔍 Optimization Opportunities
1. Build Test Suite — 2.0M tokens/run average
Token Analysis: The highest token consumer, using 9.9M tokens across 5 runs (36% of all tokens). This workflow builds and tests the entire project.
Recommendations:
Trend: Increased from 9.0M (4 runs) in previous report to 9.9M (5 runs) — one additional run but token efficiency similar.
2. Test Coverage Improver — 2.6M tokens in single run
Token Analysis: Single run consumed 2.6M tokens, suggesting extensive analysis or iteration.
Recommendations:
Trend: Increased from 2.3M in previous report — 13% increase suggests growing complexity.
3. API Surface & Export Audit — 853K tokens/run average
Token Analysis: 2.6M tokens across 3 runs. Likely scans many files for API boundaries.
Recommendations:
Trend: Increased from 2.3M (3 runs) to 2.6M (3 runs) — 13% increase per run.
4. CI/CD Pipelines Gap Assessment — 1.3M tokens
Token Analysis: Single comprehensive analysis run, slightly lower than previous report (1.5M).
Recommendations:
View Detailed Per-Workflow Analysis
Build Test Suite (5 runs, 9.9M tokens)
Runs:
Pattern: Consistent 2.0-2.4M token usage per run. All complete in 1 turn (efficient, no retries).
Impact: Highest total token consumer. Small optimizations would have significant cost impact.
Test Coverage Workflows
Test Coverage Improver (1 run, 2.6M tokens):
Test Coverage Reporter (4 runs, 2.1M tokens):
Smoke Test Workflows
Smoke tests are generally efficient, averaging 100-450K tokens/run:
These workflows are well-optimized for their purpose. No optimization needed.
Analysis & Advisory Workflows
One-time or infrequent workflows for code quality:
These are complex analysis tasks with appropriate token usage for their scope.
Workflows Without Token Data
The following workflows had runs but reported zero tokens — they may not have api-proxy instrumentation enabled:
Note: Token tracking requires the
--enable-api-proxyflag in awf configuration. These workflows may be using direct API calls or running outside the firewall.Historical Trend
Comparison to Previous Report (#3622):
Key Observations:
Trending Up 🔺:
Action Required: Monitor Build Test Suite closely — it's the largest consumer and growing. Consider optimization strategies listed above.
Key Takeaways
Recommendations
Immediate Actions:
Long-term Optimizations:
Previous Report
References: