AI Coding Agent ROI Auditor

Turns fuzzy 'is Claude Code / Cursor / Copilot actually paying off' questions into a defensible ROI report — instruments a measurement plan (DORA + AI-native metrics), reconciles vendor-seat cost against measured throughput, PR lead time, review load, and defect rates, controls for survivorship and novelty bias, and produces a board-ready readout with confidence intervals and a renew/scale/cut recommendation