There is no platform-level infrastructure to measure or surface agent sustainability metrics: compute cost versus value generated, agent survival rates, economic viability distributions across the ecosystem. Individual agents tracking their own unit economics discover severe mismatches (e.g., $1.73/day cost vs. $0.004/day value), but this data is not aggregated or visible at the ecosystem level. Without this observability layer, neither operators nor platform owners can identify systemic economic failure patterns or intervene.

AI agents fail silently because nobody aggregates cost-vs-value data across the ecosystem; operators can't benchmark viability and platforms can't spot systemic collapse patterns.

AI agent platform operators (e.g., teams running 50+ agents) and agent framework developers who need ecosystem-level economic intelligence to reduce churn and waste.

Agent operators already discover horrific unit economics ($1.73 cost / $0.004 value) only after burning money — a benchmarking layer that surfaces this BEFORE deployment turns invisible failure into a preventable decision, saving thousands per month per team.

Lightweight SDK that agents embed to report cost/value telemetry to a central index; MVP is an open-source collector + a dashboard showing anonymized ecosystem benchmarks, survival curves, and cost-efficiency percentiles — ship in 3-4 weeks using ClickHouse and a simple API.

Adjacent APM/observability market is $20B+; the agent-specific economics slice could be $500M+ as agent deployments scale to millions, with every agent framework and orchestrator as a distribution channel.