

HolmesGPT
#5 in Observability & EvalsUnknown · since Oktober 2025 (Aufnahme als CNCF Sandbox Projekt, 8. Oktober 2025) · 2× · last seen Sep 15, 2026
HolmesGPT is an open-source AI agent for automated investigation of production incidents in Kubernetes and cloud-native environments. It combines observability telemetry (logs, metrics, traces) with LLM reasoning and agentic tool calls to deliver root-cause analyses and remediation suggestions. Originally built by Robusta.dev, it now receives major contributions from Microsoft and was accepted as a CNCF Sandbox project in October 2025. It supports numerous LLM providers, operates read-only by default, and can be run via CLI, Helm/Kubernetes, or the Robusta SaaS platform.
Features
| Deployment (Self-Hosted/Cloud) | Self-hosted via CLI/Homebrew/Helm on Kubernetes, or hosted via Robusta SaaS platform (SOC 2, regions US/EU/APAC) |
| License | Apache 2.0, CNCF Sandbox project |
| Platform | Kubernetes-native (Helm deployment), CLI; also works without Kubernetes on VMs, bare metal, and cloud services |
| Price | Core project free/open source; Robusta SaaS free to start with paid tiers for larger teams; self-hosted platform requires enterprise plan |
| Protocol Compatibility | Supports Model Context Protocol (MCP) as a client (e.g. GitHub, AKS, Kubernetes MCP servers) via stdio, SSE (deprecated), and streamable-http |
| Release Date | Accepted to CNCF Sandbox on October 8, 2025 |
| Supported Models/Providers | Any LLM provider: OpenAI, Anthropic, Azure, AWS Bedrock, Google Gemini/Vertex AI, Ollama (local), and more |