Skip to main content
Glama

Start cluster

start_cluster
Destructive

Start a stopped Ankra-provisioned self-managed cluster: re-provisions its VMs from the configuration preserved at stop time (on Scaleway, AWS and Ankra Cloud, powers the preserved servers back on instead of rebuilding them) and brings Kubernetes and the installed stacks back up. Refused unless the cluster is stopped, a retry of a start whose control plane never came up, or a full start bringing parked workers back. When the stop deleted the VMs and a cluster state snapshot was captured after the cluster's last start (the stop's own capture, otherwise the newest scheduled etcd backup), the first control plane restores it (etcd reset from the snapshot, stale Node and VolumeAttachment objects cleared) instead of bootstrapping an empty cluster, so workloads come back with their Secrets, ConfigMaps, custom resources and persistent volumes; a snapshot from before the last start is restored only when named with restore_snapshot_id (the refusal names it), and a restore already requested is always kept. A paused cluster keeps its etcd and gets no new restore. The result names the snapshot (state_snapshot_id, state_snapshot_trigger, state_snapshot_captured_at) and says in its message what happens to the cluster state. restore_state=false starts fresh instead. Optional scope: 'all' (default) restores the full cluster; 'control_plane' restores only the control plane and bastion (e.g. to verify a change_control_plane result cheaply). Provisioning runs in the background as tracked operations.

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "start_cluster"
  }
}

Input Schema

NameRequiredDescriptionDefault
scopeNoWhat to start. Default: 'all'.
cluster_idNoUUID of the target cluster. Optional when a cluster is selected.
cluster_nameNoExact name of the target cluster. Optional when a cluster is selected.
restore_stateNoDefault (omitted): restore an already requested snapshot, otherwise, when the stop deleted the VMs, the newest state snapshot captured after the cluster's last start (the stop's capture or the newest scheduled backup); an older one is refused with its id; a paused cluster keeps its etcd and gets no new restore. true: the same, refusing when there is nothing to restore (a no-op on a paused cluster). false: start with an empty etcd and withdraw any pending restore (refused while an install is mid-restore); the result says how many captured snapshots are left unrestored.
restore_snapshot_idNoId of the captured state snapshot to restore, whatever its age (the id a refused start names, or one from the cluster's snapshot list). Not with restore_state=false.

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

A4.5/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations only flag destructiveHint=true and idempotentHint=false; the description goes far beyond by explaining snapshot restore logic, that stale Node/VolumeAttachment objects are cleared, that paused clusters keep their etcd, that provisioning runs as background tracked operations, and what the result message reports. This is exactly the behavioral context the annotations cannot convey.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness3/5

Is the description appropriately sized, front-loaded, and free of redundancy?

It is front-loaded correctly with the core action and scope, but the second and third sentences are extremely long run-ons packing snapshot fallback rules, refusal conditions and result fields together. The content is valuable yet dense enough to slow an agent parsing it.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a destructive, non-idempotent start with complex snapshot semantics and no output schema, the description covers preconditions, per-parameter behavior, failure modes, side effects and what the result reports. Nothing an agent needs to decide whether and how to call it is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is already 100%, so the baseline is 3, but the description adds real meaning: it explains the default restore behavior per restore_state value, the mid-restore refusal, the paused-cluster no-op, and the interaction between restore_snapshot_id and restore_state=false. The amount of edge-case semantics exceeds what the schema descriptions alone convey.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The opening sentence gives a specific verb and resource ("Start a stopped Ankra-provisioned self-managed cluster") and names the platform scope, which distinguishes it from start_managed_cluster. It also states the concrete outcome (VMs re-provisioned, Kubernetes and stacks brought back up) rather than restating the title.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It spells out the exact preconditions allowing a start (cluster is stopped, retry of a failed start, full start bringing parked workers back) and even suggests a use case for scope='control_plane' (cheaply verifying a change_control_plane result). What it lacks is a direct pointer to a sibling alternative such as start_managed_cluster for non-self-managed clusters.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.