Skip to main content
Glama

Create alert

create_alert
Destructive

Create an alert definition: a named alert with one or more rules the platform evaluates every minute. Use this when asked to create, add or set up an alert on Ankra (an Alertmanager silence is create_alertmanager_silence instead). Before calling: list_alerts to avoid a duplicate name (names are unique per organisation), list_clusters for cluster ids, and list_alert_destinations if integration_names will be set. Each rule targets resource_type 'cluster' (cluster health) or 'cluster_resource' (the stacks, addons, manifests and GitOps repositories on a cluster; cluster_resource_types is then required), is scoped by cluster_ids, environments or all_clusters=true (a rule naming none targets the chat's bound cluster), and carries a severity (critical/warning/info, default warning) and cooldown_minutes between re-fires (default 15). Condition vocabulary, by resource_type. cluster: cluster_state (condition_value offline|online, operator eq|neq); gitops_sync (failing|degraded|healthy, eq|neq); metric_threshold (condition_value is the metric key: cpu_utilisation_pct, memory_utilisation_pct, disk_utilisation_pct (cluster-wide average), node_disk_utilisation_max_pct (fullest node), pod_restarts_15m, oom_kills_15m, nodes_not_ready, nodes_disk_pressure (nodes with the kubelet DiskPressure condition), pods_pending, pods_failed (pods in phase Failed, which is where kubelet evictions land); operator gt|gte|lt|lte|eq|neq against threshold_value); metric_absence (condition_value may be omitted - it is fixed to platform_metrics; fires when the cluster has shipped no platform metrics for more than threshold_value minutes); security_findings (condition_value is the posture metric: critical, high, severe, fixable_critical, fixable_high, new_severe_cves, new_kev_cves, kev, risk_score; compared against threshold_value). cluster_resource: resource_state (all_up|up|creating|updating|stopping|down) and job_status (failed|timeout|running|success|pending|blocked|cancelled|cancelling), both eq|neq. Conditions within one rule combine with logical_operator AND (default) or OR. Any other condition type, value or operator is refused rather than stored, because the evaluator would never fire it and the alert would sit dead while reading as configured. Firings notify through the organisation's notification routes (list_notification_routes); integration_names additionally wires the alert directly to named destinations from list_alert_destinations, and an unknown name is refused. The alert is created enabled unless enabled=false. Reversible with delete_alert; to quieten it later prefer set_alert_enabled.

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": {
      "name": "<name>",
      "rules": "<rules>"
    },
    "name": "create_alert"
  }
}

Input Schema

NameRequiredDescriptionDefault
nameYesAlert name, unique within the organisation (at most 100 characters).
rulesYesThe rules evaluated for this alert; the alert fires when any rule's conditions are met.
enabledNofalse creates the alert muted. Default true.
descriptionNoWhat the alert watches and why (at most 500 characters).
integration_namesNoNames of existing alert destinations (list_alert_destinations) to notify directly, in addition to the organisation's notification routes. An unknown name is refused.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changed
    • changedInput schema / properties / rules / items / properties / min_failure_duration_minutes / description
      Previous value: -"Minutes the rule must keep reading triggered before it fires; a healthy evaluation in between resets it. Omit for the default: 5 for a pods_pending condition (a starting pod is Pending for seconds), otherwise 0, which fires on the first evaluation."New value: +"Minutes the rule must keep reading triggered before it fires; a healthy evaluation in between resets it. Omit for the default: 5 for a pods_pending or pods_failed condition (a starting pod is Pending for seconds, and a pod a rolling update deletes can be Failed for seconds), otherwise 0, which fires on the first evaluation."
  2. First observed

TDQS

A4.3/5.0
Behavior3/5

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

Annotations already declare destructiveHint=true, idempotentHint=false and readOnlyHint=false, so the mutation profile is known. The description adds real value on top: unique-per-organisation names, unknown integration names refused, invalid condition types/values/operators refused rather than silently stored, created enabled unless enabled=false, and reversibility via delete_alert. It does not, however, explain why a create carries destructiveHint=true or what downstream effects creating an alert has on notification traffic beyond naming the routes, so it is strong but not exhaustive.

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?

Purpose and usage are correctly front-loaded in the first two sentences, and nearly every clause carries information. But the middle is a single sprawling sentence enumerating the condition vocabulary as a semicolon-heavy run-on, which is hard to scan and would be far clearer as a per-resource_type list. Dense rather than wasteful, yet the formatting costs readability.

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 nested, validation-heavy mutation tool with no output schema, the description covers what an agent needs to construct a valid call: prerequisites, scoping precedence (cluster_ids vs environments vs all_clusters vs the chat's bound cluster), required companion field cluster_resource_types, defaults for severity/cooldown/min_failure_duration, and refusal behavior. Nothing material is left to inference.

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 structured fields already define every parameter and default. The description still earns credit by supplying the condition_value vocabulary that the schema leaves generic — the full metric key list (cpu_utilisation_pct, node_disk_utilisation_max_pct, oom_kills_15m, pods_failed, etc.) and the security_findings posture metrics — plus the tie between resource_type and which condition types are legal. It stops short of restating operator/threshold pairings exhaustively, which is fine given the schema.

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 concrete verb+resource+shape: "Create an alert definition: a named alert with one or more rules the platform evaluates every minute." It immediately distinguishes itself from the nearest sibling by naming create_alertmanager_silence as the silence path, so an agent can route correctly without opening any schema.

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?

Gives an explicit trigger ("Use this when asked to create, add or set up an alert"), an explicit exclusion (Alertmanager silence), and a concrete pre-call checklist: list_alerts for duplicate names, list_clusters for ids, list_alert_destinations if integration_names is set, list_notification_routes for the default route. It even routes the follow-up case: "to quieten it later prefer set_alert_enabled."

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.