HolmesGPT

A self-hosted SRE AI agent investigates production incidents using LLM APIs and runs scheduled health checks through a Kubernetes operator.

HolmesGPT is a self-hosted AI agent for DevOps and SRE teams investigating production incidents. It queries live observability data to find root causes across Kubernetes, VMs, bare metal, cloud services and databases.

Incident investigation: Connect Prometheus, Grafana, Datadog, cloud diagnostics and private Confluence runbooks, with support for additional data sources through REST APIs.
Alert follow-up: Fetch alerts or tickets from AlertManager, PagerDuty, OpsGenie and Jira, then write findings back to the originating system; PagerDuty, OpsGenie and Jira integrations require the CLI.
Background checks: Operator mode performs scheduled service checks and deployment verification, sends findings and proposed fixes to Slack, and can open pull requests through a GitHub integration.
Interactive troubleshooting: Ask questions and follow-ups through a terminal, web interface or K9s plugin.

The agent is open source under Apache 2.0 and runs on your infrastructure. General investigations don't require Kubernetes, but operator mode does. Its health checks can query connected systems outside the cluster.

Model connections use API credentials for providers including OpenAI, Anthropic, Azure, Bedrock and Gemini. Requests to these cloud providers leave your infrastructure even though the agent is self-hosted. Observability access is read-only and respects RBAC permissions. The agent filters large query results before passing them to the model and limits tool output and memory use.