[feorlen]

Content description

Rebuilds the vCluster integrations overview with the card layout from Start here, grouped by reader goal, so readers can find integrations across both the vCluster and Platform docs.

• Card sections: GitOps and CI/CD, GPU scheduling and inference, workload identity, secrets and certificates, networking, and infrastructure and fleet management. 25 cards cover every vCluster integration page plus the related Platform integrations: Argo CD and Akuity, GitLab CI, Vault, Karpenter, Certified Stacks, NVIDIA Run:ai, NVIDIA Dynamo, and the KubeVirt node provider.
• Card tags show the worker node models (Shared nodes, Private nodes), the license tier (Free, Enterprise), and "vCluster Platform" for Platform pages. Each tag matches its destination page's tenancy notice and sidebar badge.
• KubeVirt has two clearly named cards: "VMs as workloads" (vCluster) and "VMs as nodes" (Platform node provider).
• Intro keeps the three ways integrations work and links to the Platform docs for integrations not listed. Each card section opens with a one-sentence introduction.
• Untagged cards: tags repeat what each guide states, so a card without a tag means its guide doesn't state that requirement. The intro says so.
• Metrics server stays under "reuse on the control plane cluster", and its card covers both methods its guide documents.
• LinkCard gets an optional tags prop. Existing cards are unchanged.

Platform links on next: Platform cards use /docs/platform/next/... paths. Until the release, this build's stable Platform version is 4.12, where the 5.0 paths don't exist, so product="platform" links would fail the build. An MDX comment at the top of the page marks the release-day switch to product="platform".

KubeVirt node provider badge: the "KubeVirt: VMs as nodes" card is tagged Free, matching its destination page. This was confirmed against the plan definitions (details in DOC-1974). The Enterprise badge on the vMetal virtual machines overview covers a separate feature, so it doesn't conflict.

Testing: the Netlify deploy preview builds, validate-mdx-links passes, and Vale has one warning left (the "Networking" heading, which names a group of cards). A local production build confirmed every card link resolves. In the browser, the grid is 3 columns on desktop and 1 column at narrow width, card clicks across products land on the right pages, and Tab and Enter work with a visible focus ring.

Preview link

https://deploy-preview-3126--next-vcluster-docs-site.netlify.app/docs/vcluster/next/integrations

Internal reference

Closes DOC-1974

Follow-ups from the AI review: DOC-1977 (metrics server guide order) and DOC-1978 (preview environments platform requirement).

AI review: mention @claude in a comment to request a review. Claude reads the PR and posts review comments. See CONTRIBUTING.md for example commands.

Only members of the loft-sh organization can trigger @claude. If you're contributing from outside vCluster Labs, ask a maintainer to request the review.

🤖 Generated with Claude Code

@netlify /docs

[netlify]

✅ Deploy Preview for next-vcluster-docs-site ready!

Name	Link
🔨 Latest commit	431257b
🔍 Latest deploy log	https://app.netlify.com/projects/next-vcluster-docs-site/deploys/6af64f7f15999ee2a712bedb
😎 Deploy Preview	https://deploy-preview-3126--next-vcluster-docs-site.netlify.app/docs
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[feorlen]

@claude review this PR

[claude]

Claude finished @feorlen's task in 1m 11s —— View job

---

PR review: rebuild the integrations overview with cards

• Read the diff
• Check card tags against each destination page's tenancy notice and sidebar badge
• Check link targets and component code
• Post review

This PR is in good shape. I checked every card tag against the TenancySupport notice and sidebar_class_name / _category_.json badge on its destination page, and they all match:

• vCluster pages: tiers come from pro and free, including the free category on pod-identity. Node models come from hostNodes and privateNodes. setup-vind has no notice, so it has no tag.
• Platform pages: Argo CD overview is free, Vault is pro, and the KubeVirt driver page is free. GitLab CI, Karpenter, Certified Stacks, Run:ai, and Dynamo have no badge, so no tier tag is right.

Every Platform to path points at a page that exists on this branch. The Argo CD path resolves through the slug: in argocd/overview.mdx, and Certified Stacks resolves through index.mdx. The LinkCard change is backward compatible, and the tags?.length > 0 check stops a stray 0 from rendering.

Should fix

1. The metrics server move only half matches its guide. vcluster/integrations/overview.mdx, intro bullets and the "Metrics server" card.

vcluster/integrations/metrics-server.mdx covers two methods:

• Installing a dedicated metrics server inside the cluster (lines 13-26).
• "Reuse host metrics server" through integrations.metricsServer, pulled in from _fragments/metrics-server.mdx at line 60.

Line 15 of the guide calls the reuse method the recommended one. It reads: "In case the above recommended method of getting metrics in vCluster using the metrics server proxy does not fulfill your requirements…". So the original placement under "Reuse a tool on the control plane cluster" was arguably right, and the card text ("Install a dedicated metrics server…") describes only the fallback. Two options:

• Keep the new bullet, but change the card so it covers both methods. For example: "Make kubectl top work in a cluster, either by reusing the control plane cluster's metrics server or by installing a dedicated one."
• Or put the metrics server back in the "reuse" bullet and drop "a dedicated metrics server" from the "run inside" bullet.

The guide also describes the recommended method as "above" when it's actually at the bottom of the page. That's an existing problem and out of scope here, but worth a follow-up ticket.

Fix this →

Consider

2. Preview environments needs the platform, but the card has no tag saying so. The card description says "with the platform's sleep mode", and the guide creates a cluster template in vCluster Platform. The page has no tier badge, so by your rule ("tags repeat what each guide states") having no tier tag is consistent. Still, a reader who filters on tags will take this for a plain vCluster guide. This probably needs a badge on the destination page, not just a tag on the card. That could be a follow-up.

3. "idle environments use no resources" overstates what sleep mode does. The guide says the same thing (line 14). Sleep mode scales workloads down, but persistent volumes and the cluster's stored state remain. Something like "so idle environments stop using compute" is more accurate. It also fits the repositioning guidance in CLAUDE.md to move away from cost framing.

4. Tag text is part of each link's accessible name. Screen readers will read a card as "Flux Shared nodes Private nodes Manage clusters as code…". That's acceptable. If you want the tags announced as a separate unit, add a visually hidden prefix such as "Supports:" or wrap the tags in <span role="list"> / role="listitem". This is optional.

Not an issue

• The /docs/platform/next/... hardcoding is justified, and the MDX comment documents the release-day switch to product="platform". Keeping that comment is the right call.
• The "Networking" heading Vale warning is fine. It names a group of cards.
• Terminology: "cluster", "control plane cluster", and "tenant" follow the repositioning rules, and no "virtual cluster" or "tenant cluster" slipped into the prose.
• The KubeVirt Free tag doesn't conflict with the Enterprise badge on the vMetal VM overview page. You covered that in the PR body.
• Rancher card: the guide confirms automatic registration and cluster owner grants (rancher.mdx:76).

Not run: I didn't run a local build or Vale. I'm relying on the Netlify preview and the checks listed in the PR body.