Ingero

Ingero is an MCP (Model Context Protocol) server that enables AI assistants to perform advanced, zero-intrusion debugging and observability on GPU workloads. Leveraging eBPF, it traces the full causal chain from Linux kernel and CUDA driver/runtime events through to Python source code, correlating system, kernel, and CUDA-layer events for root-cause analysis of GPU slowdowns, stalls, or failures. Ingero directly integrates with CUDA (runtime and driver), NVIDIA hardware, Python ML stacks (such as PyTorch), and Kubernetes environments. It is aimed at site reliability engineers (SREs), AI/ML practitioners, devops, and anyone running production machine learning or GPU-heavy applications who need automated, in-depth GPU debugging and performance root-cause analysis—accessible to AIs via MCP tools rather than only humans.