Documentation
Install it, scope it,
then read the evidence.
These are the same guides we'd hand a new SRE on the team: what OpsKit reads, what it refuses to touch, and exactly what evidence it gathers for the failure modes that page people at 2am.
Install OpsKitInstall OpsKitInstall OpsKit's read-only Kubernetes MCP server, 42 specialist agents, and safety hooks in one command. Requirements, what gets written to disk, and how to verify the install.AI Agent Kubernetes SecurityKubernetes AI agent securityHow to let an AI coding agent read a production Kubernetes cluster without risking a mutation. RBAC scoping, the opskit-guard hook allowlist, and what to audit before your first pilot.CrashLoopBackOff DiagnosisCrashLoopBackOff AI diagnosisHow OpsKit's k8s-investigator agent diagnoses a CrashLoopBackOff: which evidence it gathers, how it forms a hypothesis, and how the fix reaches you as a GitOps PR instead of a live mutation.OOMKilled InvestigationOOMKilled Kubernetes investigationDiagnosing OOMKilled pods with an AI agent that reads resource usage, limits, and events directly from the cluster instead of guessing at a memory leak.Pending Pod DiagnosisPending pod Kubernetes diagnosisWhy pods get stuck in Pending — taints, resource pressure, PVC binding, admission policy — and how OpsKit's agents attribute the cause before proposing a fix.GitOps PR Evidence ReplayGitOps PR evidence replayHow opskit evidence bundle attaches the exact redacted MCP tool output an agent used to a GitOps pull request, so reviewers see the evidence, not just a summary.Namespace-Scoped Read AccessNamespace-scoped Kubernetes read access for AI agentsHow to limit an AI agent's cluster reads to an allowlist of namespaces and labels, enforced at the Kubernetes RBAC layer instead of only in a prompt.