Rancher MCP

List certificates as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cis_scan and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Fetch one cis_scan_profile and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cis_scan_profiles as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

List cis_scans as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cloud_credential and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cloud_credentials as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cluster_alert_rule and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_alert_rules as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cluster_driver and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_drivers as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

List Kubernetes events across the whole cluster by default — the fastest way to see "what just happened here" after an incident. Pass `namespace` to scope to one namespace instead. Filter by event_type (Warning/Normal) or reason.

Fetch one cluster_flow and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_flows as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cluster and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Diagnose one cluster in a single call: state, condition failures, unhealthy components, and a node-readiness rollup, plus a derived list of concrete issues — prefer this over separately fetching cluster and node details.

Fetch one cluster_logging and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_loggings as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Roll up one cluster's nodes into ready, not-ready, and unschedulable counts and report the tallies — a lighter-weight alternative to a full node listing when only the totals matter.

Fetch one cluster_output and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_outputs as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cluster_policy_report and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_policy_reports as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cluster_registration_token and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_registration_tokens as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one cluster_role_template_binding and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List cluster_role_template_bindings as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Summarize health across every cluster on an instance in one call — per-cluster state, node readiness, top issues, and a fleet-wide severity breakdown — the fastest way to triage an entire estate.

List clusters as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Apply one config_map to a desired state as a declarative PUT and return its typed detail — creating the object when absent, replacing it when present. This writes to the cluster.

Create one config_map from the supplied spec and return the typed detail of the new object, including its server-assigned identity and initial state. This writes to the cluster.

Fetch one config_map and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List config_maps as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one cron_job and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Modify one cron_job in place (resume) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

Modify one cron_job in place (suspend) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

List cron_jobs as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one daemonset and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List daemonsets as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one deployment and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Modify one deployment in place (pause) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

Modify one deployment in place (restart) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

Modify one deployment in place (resume) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

Modify one deployment in place (scale) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

List deployments as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one endpoint_slice and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List endpoint_slices as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one etcd_backup and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List etcd_backups as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one feature and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List features as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Scan for pods stuck in CrashLoopBackOff, ImagePullBackOff, Pending, Failed, or Running-but-not-Ready, and report each with its reason, restart count, and owning controller. Omit `namespace` to sweep the entire cluster in one call instead of one namespace at a time.

Scan pod disruption budgets that currently allow zero voluntary evictions and report each with its healthy and desired counts plus the blocking reason — explains why a drain or rollout is stuck waiting. Omit `namespace` to sweep the entire cluster in one call.

Scan services that have a selector but no matching endpoint addresses — a fast signal that traffic is arriving nowhere — reporting each service's type and selector. Omit `namespace` to sweep the entire cluster in one call.

Scan deployments and statefulsets for replica counts that have not converged, reporting each stalled rollout's desired, ready, and updated counts plus the stuck reason. Omit `namespace` to sweep the entire cluster instead of one namespace at a time.

Scan persistent volume claims stuck outside the Bound phase and report each with its storage class, requested size, and stall reason — the quickest way to spot storage blocking a pod from starting. Omit `namespace` to sweep the entire cluster at once.

Identify nodes that are not Ready or have been marked unschedulable, reporting each with its roles, reason, and how long the condition has held. Omit `cluster_id` to sweep every cluster on the instance at once.

Fetch one fleet_workspace and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List fleet_workspaces as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one flow and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List flows as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one global_role_binding and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List global_role_bindings as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one global_role and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List global_roles as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one group and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List groups as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one horizontal_pod_autoscaler and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Modify one horizontal_pod_autoscaler in place (set min max) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

List horizontal_pod_autoscalers as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one ingress and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List ingresses as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

List every Rancher instance this deployment knows about, each with its base URL and target version, so an agent can pick a valid `instance` argument before calling any other tool.

Fetch one job and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List jobs as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one limit_range and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List limit_ranges as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one longhorn_backup and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List longhorn_backups as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one longhorn_node and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List longhorn_nodes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one longhorn_snapshot and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List longhorn_snapshots as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one longhorn_volume and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List longhorn_volumes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Check if Rancher monitoring is enabled on a cluster and summarize its state.

Fetch one namespace and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Aggregate one namespace's pod-phase tallies and deployment, daemonset, and statefulset readiness into a single rollup, saving the four separate calls otherwise needed to answer "is this namespace healthy".

Fetch one namespaced_certificate and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List namespaced_certificates as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

List namespaces as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

List network_policies as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one network_policy and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Mark a Rancher-managed node unschedulable so the scheduler stops placing new pods there, without evicting anything already running, and return the action result — reversible with `rancher_node_uncordon`.

Fetch one node_driver and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List node_drivers as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one node and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Fetch one node_template and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List node_templates as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Restore scheduling on a previously cordoned Rancher-managed node so the scheduler resumes placing new pods there, and return the action result.

List nodes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Invoke a named Norman resource action from that resource's own `actions` links, with an optional JSON body, and return its response — the escape hatch for lifecycle operations no curated tool wraps yet.

Replace one Norman resource's full body via PUT under any schema id and return the updated object — the untyped escape hatch for updates no curated apply tool yet covers.

Create a new Norman resource from a raw JSON payload under any schema id and return the resulting object — the untyped escape hatch for resource types with no curated create tool yet.

Delete one Norman resource by schema id and resource id after an explicit confirmation string, and report the deletion result — the untyped, destructive escape hatch for resource types with no curated delete tool.

Fetch one Norman resource's full untyped payload by schema id and resource id — the escape hatch for resource types with no curated `rancher_*_get` tool yet.

Follow a named Norman resource link from that resource's own `links` map and return the linked payload — the escape hatch for related data no curated tool exposes yet.

List any Norman resource collection by raw schema id with full untyped payloads and server-side filtering — the escape hatch for resource types with no curated `rancher_*_list` tool yet.

Merge a partial JSON payload into one existing Norman resource under any schema id and return the patched object — the untyped escape hatch for a scoped field update with no curated patch tool.

Fetch one Norman schema's field definitions, required attributes, and supported collection actions by schema id, for constructing a valid create, apply, or patch payload.

List every Norman resource schema Rancher exposes on the management plane, identifying which type ids are valid for `rancher_norman_resource_list` and its siblings.

Fetch one notifier and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List notifiers as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one output and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List outputs as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one persistent_volume_claim and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Modify one persistent_volume_claim in place (set size) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

List persistent_volume_claims as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one persistent_volume and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List persistent_volumes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one pod_disruption_budget and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List pod_disruption_budgets as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one pod and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Fetch a pod container's recent log tail — the missing `kubectl logs` verb. Omit `container` for a single-container pod; a multi-container pod without one returns a clean error listing the available container names so you can retry. Set `previous=true` to read the last terminated instance's logs (diagnosing a crash loop) instead of the current one.

Fetch one pod_monitor and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List pod_monitors as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

List pods as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one policy_report and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List policy_reports as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one priority_class and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List priority_classes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one project and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Aggregate one project's namespaces, pod counts, failing-pod tallies, and unhealthy-workload counts into a single rollup, saving the per-namespace calls otherwise needed to answer "is this project healthy".

Fetch one project_logging and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List project_loggings as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one project_role_template_binding and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List project_role_template_bindings as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

List projects as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one prometheus_rule and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List prometheus_rules as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one replica_set and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List replica_sets as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Delete one resource of any deletable kind and return a typed receipt of what was removed. Destructive and irreversible. `resource_kind` selects the resource family (pod, deployment, config_map, secret, ...); `name` is that resource's own name (uniform across every kind). `namespace` is required for namespaced kinds, ignored for cluster-scoped ones. The caller MUST first echo the exact confirmation phrase this kind requires: `"delete <kind> <name>"`, plus `" in namespace <namespace>"` for namespaced kinds — call this tool without a matching `confirmation` first to see the required phrase in the…

List recent Kubernetes events for one named resource (Pod, Deployment, PersistentVolumeClaim, ...) — "what just happened to this thing", generalized from `pod_get`'s inlined pod-scoped events to any `kind`. Most recent first, capped to the last ~20.

Fetch one resource_quota and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List resource_quotas as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Replace metadata.annotations on one resource of any annotatable kind via JSON merge-patch. The `annotations` arg is a complete replacement map — pass {} to remove all annotations. `resource_kind` selects the resource family; `name` is that resource's own name (uniform across every kind). `namespace` is required for namespaced kinds and ignored for cluster-scoped ones. Returns a compact mutation receipt. Subject to write rate limiting and audit logging.

Replace metadata.labels on one resource of any labelable kind via JSON merge-patch. The `labels` arg is a complete replacement map — pass {} to remove all labels. `resource_kind` selects the resource family (pod, deployment, config_map, secret, ...); `name` is that resource's own name (uniform across every kind — no more per-family `pod_name` / `service_name` / `volume_name`). `namespace` is required for namespaced kinds and ignored for cluster-scoped ones (e.g. storage_class, priority_class). Returns a compact mutation receipt — the before/after of just the labels field plus timing, not the f…

Fetch one restore and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List restores as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one role_template and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List role_templates as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one runtime_class and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List runtime_classes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Create one secret from the supplied spec and return the typed detail of the new object, including its server-assigned identity and initial state. This writes to the cluster.

Fetch one secret and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect. Sensitive values are withheld by default (only names/counts appear); pass reveal=true to receive the decoded values — each reveal is audited.

List secrets as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Check the target instance's `/healthz` endpoint and report whether the Rancher management server is reachable and responding — a first liveness probe before deeper diagnosis.

Return static deployment metadata for one Rancher instance: its configured URL, primary target version, and compatibility floor, without making a network call.

Fetch the Rancher server's reported version alongside this MCP server's own build version, so an agent can confirm which API surface and capability set it is talking to.

Fetch one service_account and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List service_accounts as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one service and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Fetch one service_monitor and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List service_monitors as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Modify one service in place (set type) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

List services as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Fetch one setting and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List settings as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one statefulset and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Modify one statefulset in place (scale) via a JSON merge-patch and return a mutation receipt — the before and after of only the changed fields, not the whole object. A targeted write.

List statefulsets as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool. Omit `namespace` to list across the whole cluster; pass it to scope to one namespace.

Invoke a named Steve Kubernetes-proxy resource action with an optional JSON body and return its response — the escape hatch for lifecycle operations no curated tool wraps yet.

Apply one Steve Kubernetes-proxy resource via Kubernetes server-side apply under any schema id and return the updated object — the untyped escape hatch for updates with no curated apply tool yet.

Create a new Steve Kubernetes-proxy resource from a raw JSON payload under any schema id and return the resulting object — the untyped escape hatch when no curated create tool exists for that kind.

Delete one Steve Kubernetes-proxy resource by schema id and resource id after an explicit confirmation string, and report the deletion result — the untyped, destructive escape hatch for kinds with no curated delete tool.

Fetch one Steve Kubernetes-proxy resource's full untyped payload by schema id, cluster, and resource id — the escape hatch for kinds with no curated `rancher_*_get` tool yet.

Follow a named Steve Kubernetes-proxy resource link and return the linked payload — the escape hatch for related data no curated tool exposes yet.

List any Steve Kubernetes-proxy resource collection by raw schema id, cluster, and optional namespace with label or field selectors — the escape hatch for kinds with no curated `rancher_*_list` tool yet.

Merge a JSON merge-patch into one existing Steve Kubernetes-proxy resource under any schema id and return the patched object — the untyped escape hatch for a scoped field update with no curated patch tool.

Stream change events for a Steve Kubernetes-proxy resource collection up to a bounded event count and return them as one batch — the escape hatch for watching kinds with no curated tool yet.

Fetch one Steve schema's field definitions and supported verbs for a target cluster, for constructing a valid create, apply, patch, or watch payload by schema id.

List every Steve Kubernetes-proxy schema available for a target cluster, identifying which type ids are valid for `rancher_steve_resource_list` and its siblings.

Fetch one storage_class and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List storage_classes as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one template and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

Fetch one template_version and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List template_versions as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

List templates as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.

Fetch one user and return its full typed detail: the conditions, diagnostics, and derived fields the list summary leaves out. Call this once a list has identified which one to inspect.

List users as lightweight typed summaries — identity, state, and a per-item health rollup rather than full specs — so an agent can enumerate what exists before opening any one in detail with the matching get tool.