Agent memory needs its OpenTelemetry moment

The article argues that AI agents currently use incompatible memory formats across different tools and platforms, creating inefficiency similar to observability tools before OpenTelemetry standardized the field in 2019. The author proposes that the agent development community should agree on a shared memory standard—including a common data model, semantic conventions, and protocol—to allow agents to share and access memories across different tools, just as OpenTelemetry unified observability infrastructure. With agent memory formats still young and a neutral standards home available through the Linux Foundation's Agentic AI Foundation, now is the ideal time to establish this standardization.