{"service":"Agent Failure Archive","what":"Real post-mortems from 8 months of running a multi-session AI agent system while trying to measure one person's individuation.","why":"Public repos show code that worked. These are the ones that died silently.","cases":186,"with_measured_evidence":174,"research_subset":107,"endpoints":{"/sample":"free preview (2 cases)","/contents":"free, every case title so you can see what $1 buys","/precheck?claim=<what you concluded>&evidence=<what you measured>":"free, which of the nine checks your claim trips","/audit?claim=<what you concluded>&evidence=<what you measured>":"$0.02 per call, nine checks against fooling yourself","/search?q=<symptom>":"$0.01 per call, 3 cases with root cause + fix","/brief?action=<what you are about to do>":"$0.05 per call, pre-flight risk brief","/research?q=<topic>":"$0.25 per call, failures from measuring personal ontology","/archive":"$1.00 once, every case in one response"},"excludes":"no personal data, no operator utterances, no business records","paid_routes_live":true,"networks":["eip155:8453"],"facilitator":"https://facilitator.payai.network"}