Debugging production incidents is often more challenging than actually fixing the problems. Engineers frequently face issues like missing documentation, tool overload, and the complexity of modern systems like Kubernetes. They often spend significant time trying to piece together clues and data, which is both mentally exhausting and inefficient. HolmesGPT, a new open-source AI troubleshooting agent, aims to streamline this process by integrating various data sources and offering clear, actionable insights. HolmesGPT combines observability telemetry with large language model (LLM) reasoning to facilitate faster root cause analysis in cloud-native environments. Its design allows it to actively fetch relevant data, execute targeted queries, and refine its hypotheses based on the findings. When users input a query about a specific issue, like a pod in a crash loop, HolmesGPT breaks down the problem, queries necessary data, and returns a natural language explanation along with suggested fixes. The architecture of HolmesGPT is flexible, enabling contributors to add new toolsets and custom commands to enhance functionality. Installation can be done easily via pip or other methods, and it works with existing observability tools. The project encourages community participation, inviting users to contribute by developing integrations, creating runbooks, or improving documentation. HolmesGPT represents a significant step towards reducing the chaos of production debugging in cloud-native systems.