Groundcover has expanded its AI Observability platform to address visibility gaps in multi-step agentic AI workflows. Traditional observability tools built for deterministic software break down when dealing with long-running agent sessions involving thousands of tool calls. Groundcover's approach uses a patented eBPF sensor that captures LLM interactions at the kernel level without any code instrumentation, keeping all data within the customer's own cloud (BYOC). New capabilities include complete agent execution traces, span-level token cost tracking (distinguishing regular, cache creation, and cache read tokens), and native support for Google Vertex AI. The platform correlates LLM behavior with infrastructure telemetry, enabling teams to diagnose issues like token cost spikes caused by memory-starved Kubernetes nodes driving retries.
Sort: