HolmesGPT

AI troubleshooting agent for Kubernetes and observability data; can run as or call MCP servers.

By defaultCan write
Safest setupNo read-only mode Use a read-only credential
Flagsegress

Note: Can write (GitHub integration)

Labels are provisional (from the project README/docs, checked Oct 27, 2026). Not yet tested in our sandbox.

Read-only switch

README: "By design, HolmesGPT has read-only access and respects RBAC". The same README says its GitHub integration "can even open PRs", so not every integration is read-only.

Preset

No preset yet. We haven't confirmed an install command for this server. Start with the credential layer below: it's your only control here.

Credential layer

Bind the kubeconfig/service account the agent uses to a read-only ClusterRole (e.g. the built-in view role). view excludes Secrets; do not grant get secrets to agents.

Gotchas

  • Read-only claim covers its cluster access; the GitHub integration can open PRs. Do not give it a write-scoped GitHub token.

Facts

Vendor
HolmesGPT (CNCF sandbox; Robusta)
Source
HolmesGPT/holmesgpt
License
Apache-2.0
GitHub stars
3.7k
Latest release
0.45.0 (Oct 15, 2026)
Last push
Oct 27, 2026
Status
Active
Last checked
Oct 27, 2026

Sources

Spotted something wrong? Labels change when vendors change. Suggest a correction

Read-only alternatives in Kubernetes