Skip to main content
Glama

Create power schedule

create_power_schedule
Destructive

Create a power schedule on an Ankra-provisioned self-managed cluster: a scheduled stop or start that fires once at run_at (schedule_kind 'once') or repeatedly per a standard 5-field cron_expression evaluated in an IANA timezone (schedule_kind 'cron'), executed within a minute of falling due. By default (stop_mode 'delete_resources') a fired stop behaves exactly like stop_cluster - it TERMINATES every VM of the cluster at the provider and preserves only the configuration for a later start (except on Scaleway, AWS and Ankra Cloud, where the servers are powered off and kept with their disks, exactly like stop_cluster there) - so state the schedule's action, stop mode, and timing to the user before creating it. stop_mode 'scale_to_zero' instead DELETES the worker servers at the provider (workloads stop) and keeps the control plane (with etcd), networks, SSH keys, and cloud volumes, which keep billing; the matching start creates new workers with the previous configuration. Anything on a worker's own disk - local-path, hostPath or other node-local persistent volumes, emptyDir - is lost at every stop, so tell the user this before creating a scale_to_zero schedule, and the result's node_local_storage names the claims at risk when the cluster inventory shows any (or says it could not tell). An enabled scale_to_zero stop on a cluster whose inventory shows such claims is refused unless accept_node_local_data_loss is true: the refusal names the claims, so show them to the user, ask whether they accept losing that data at every stop, and call create_power_schedule again with accept_node_local_data_loss true only after the user has explicitly agreed. To stop a cluster at night or at weekends to save cost, use propose_off_hours_schedule instead: it files the off-hours decision with its gate, receipt and measurement. Use create_power_schedule for one-off stops and starts and for schedules the user asks for that are not the off-hours saving. A stop fired while the cluster is already stopped is skipped, a start fires on a stopped cluster (full re-provision) or on a running cluster whose workers a scale-to-zero stop removed (worker restore), and a busy cluster defers the action (retried every 5 minutes for up to 24 hours). At most 20 schedules per cluster. The schedule itself is reversible: delete_power_schedule removes it before it fires.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "action": "<action>",
      "schedule_kind": "<schedule_kind>"
    },
    "name": "create_power_schedule"
  }
}

Input Schema

NameRequiredDescriptionDefault
actionYesWhat the schedule does when it fires.
run_atNoRFC 3339 timestamp of the single fire, e.g. '2026-01-02T19:00:00Z'. Required for schedule_kind 'once'; must be in the future.
enabledNoWhether the schedule is armed. Defaults to true; pass false to create it paused.
timezoneNoIANA timezone the cron expression is evaluated in, e.g. 'Europe/Stockholm'. Defaults to UTC.
stop_modeNoHow a stop fires: 'delete_resources' (default) is the full stop_cluster teardown that terminates every VM; 'scale_to_zero' deletes the worker servers (data on their own disks is lost at every stop) and keeps the control plane, networks, SSH keys and cloud volumes; 'pause' powers every server off and keeps it with its disks (etcd and local data survive) - k3s clusters on Hetzner, UpCloud and DigitalOcean only, and what a stop always does on AWS, Scaleway and Ankra Cloud. A pause saves no compute cost on Hetzner, DigitalOcean or UpCloud (Developer and General Purpose plans): they bill a powered-off server at its full price, so say so and offer scale_to_zero or delete_resources when the user wants to save. On AWS and Scaleway a powered-off server stops billing compute; its disks and IPs keep billing. Only meaningful with action 'stop'.
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.
schedule_kindYes'once' fires a single time at run_at; 'cron' fires repeatedly per cron_expression.
preserve_stateNoFor a 'stop' schedule with stop_mode 'delete_resources' on a provider whose stop deletes the VMs: capture the cluster's state (an encrypted etcd snapshot) before the teardown, and restore it on the next start. When the capture cannot run, the stop is refused and nothing is torn down; the failure says to set preserve_state false to allow a stop without saving state. Defaults to true on create, so the stop captures state unless you pass false; on update an omitted value keeps the schedule's current setting. Ignored for other schedules.
cron_expressionNoStandard 5-field cron expression, e.g. '0 19 * * 1-5'. Required for schedule_kind 'cron'. Must not embed a timezone; use the timezone parameter.
accept_node_local_data_lossNoThe user's explicit consent to losing the data on the worker disks at every stop of an enabled scale_to_zero stop schedule. On a cluster whose inventory shows persistent volume claims on a worker's own disk, such a schedule is refused without it and the refusal names the claims. Set it to true ONLY after you have shown the user those claims and the user has explicitly agreed to lose that data - never on your own judgement, never pre-emptively, and not because the user asked for the schedule. Ignored for every other schedule; defaults to false.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changed
    • changedInput schema / properties / stop_mode / description
      Previous value: -"How a stop fires: 'delete_resources' (default) is the full stop_cluster teardown that terminates every VM; 'scale_to_zero' deletes the worker servers (data on their own disks is lost at every stop) and keeps the control plane, networks, SSH keys and cloud volumes; 'pause' powers every server off and keeps it with its disks (etcd and local data survive) - k3s clusters on Hetzner, UpCloud and DigitalOcean only, and what a stop always does on AWS and Scaleway. A pause saves no compute cost on Hetzner, DigitalOcean or UpCloud (Developer and General Purpose plans): they bill a powered-off server at its full price, so say so and offer scale_to_zero or delete_resources when the user wants to save. On AWS and Scaleway a powered-off server stops billing compute; its disks and IPs keep billing. Only meaningful with action 'stop'."New value: +"How a stop fires: 'delete_resources' (default) is the full stop_cluster teardown that terminates every VM; 'scale_to_zero' deletes the worker servers (data on their own disks is lost at every stop) and keeps the control plane, networks, SSH keys and cloud volumes; 'pause' powers every server off and keeps it with its disks (etcd and local data survive) - k3s clusters on Hetzner, UpCloud and DigitalOcean only, and what a stop always does on AWS, Scaleway and Ankra Cloud. A pause saves no compute cost on Hetzner, DigitalOcean or UpCloud (Developer and General Purpose plans): they bill a powered-off server at its full price, so say so and offer scale_to_zero or delete_resources when the user wants to save. On AWS and Scaleway a powered-off server stops billing compute; its disks and IPs keep billing. Only meaningful with action 'stop'."
  2. Changed1 schema field changed
    • changedInput schema / properties / preserve_state / description
      Previous value: -"For a 'stop' schedule with stop_mode 'delete_resources' on a provider whose stop deletes the VMs: capture the cluster's state (an encrypted etcd snapshot) before the teardown, and restore it on the next start. When the capture cannot run, the stop is refused and nothing is torn down; the failure says to set preserve_state false to allow a stop without saving state. Defaults to false on create; on update an omitted value keeps the schedule's current setting. Ignored for other schedules."New value: +"For a 'stop' schedule with stop_mode 'delete_resources' on a provider whose stop deletes the VMs: capture the cluster's state (an encrypted etcd snapshot) before the teardown, and restore it on the next start. When the capture cannot run, the stop is refused and nothing is torn down; the failure says to set preserve_state false to allow a stop without saving state. Defaults to true on create, so the stop captures state unless you pass false; on update an omitted value keeps the schedule's current setting. Ignored for other schedules."
  3. First observed

TDQS

A4.7/5.0
Behavior5/5

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

Annotations only say destructive=true; the description goes far beyond, spelling out provider-specific semantics (VM termination vs. power-off on Scaleway/AWS/Ankra Cloud), billing consequences of pause, exact data-loss surface (local-path, hostPath, emptyDir, node-local PVs), refusal conditions, the 20-schedule cap, skip/defer/retry behaviour (5 minutes for up to 24 hours), and firing latency within a minute.

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?

Content is largely load-bearing safety information, but it is delivered as one dense paragraph of roughly 350 words with no headings or line breaks, making provider-specific and consent rules hard to scan. Several clauses (e.g. the repeated restatement of what delete_resources does and the parenthetical provider lists) could be tightened without loss.

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, 11-parameter scheduling tool with no output schema, the description covers the action/timing model, data-loss consequences, consent gate, refusal behaviour, limits, and even what the result's node_local_storage field reports. An agent has everything needed to create the schedule correctly and warn the user beforehand.

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 100%, so the baseline is 3, but the description adds cross-parameter meaning the schema does not synthesize: the interaction between enabled + scale_to_zero + inventory claims that triggers refusal, and the requirement that accept_node_local_data_loss be set only after explicit user agreement. It does not add much beyond the (already rich) schema for run_at/cron_expression/timezone, so not a 5.

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?

Opens with a specific verb and resource ("Create a power schedule on an Ankra-provisioned self-managed cluster") and immediately defines the two schedule kinds and what a fired schedule does. It is clearly separable from stop_cluster, propose_off_hours_schedule, and delete_power_schedule, each of which is named with a distinct role.

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

Usage Guidelines5/5

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

Explicitly routes the agent: for night/weekend cost saving use propose_off_hours_schedule, while create_power_schedule is for one-off stops/starts and non-off-hours schedules. It also gives the reversal path (delete_power_schedule) and an explicit consent protocol (show claims, ask the user, only then retry with accept_node_local_data_loss true).

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.