Skip to main content
Glama
Sign In

Costory: Your Finops MCP

by io.costory.app-api
OverviewDiscussionsAdmin

preview_alert

Read-only

Backtest a cost-alert condition BEFORE creating it: replays the condition against the last lookbackDays (default 45) of data and reports how many times it would have fired. Takes the same queries + condition + dedup as create_alert (no notification channel needed). The dedup window is per groupBy value; delivery stays one message listing newly eligible groups. Returns the evaluation window, firingDays (distinct days the condition held), firingRows (per-group fires), notificationsCount (days on which at least one newly eligible group would notify) and a sample of firing dates with per-group state. Use this to sanity-check a condition/threshold (and tune dedup) before calling create_alert. EXAMPLE: "Would 'alert if 7-day AWS spend tops $50k' have fired this month?" → { queries: [{ type: "cost", name: "a", metricId: "cost", currency: "USD", filterCel: "cos_provider in ["AWS"]" }], condition: "rollingSum(a, 7, DAY) > 50000", dedup: { kind: "CALENDAR", calendarUnit: "WEEK" }, lookbackDays: 30 }

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": {
      "queries": "<queries>",
      "condition": "<condition>",
      "context": "<context>"
    },
    "name": "preview_alert"
  }
}

Input Schema

NameRequiredDescriptionDefault
slugNoOrganization slug. Omit to auto-detect from your account (fails if you belong to multiple orgs).
dedupNoDeduplication config controlling how often a still-firing group notifies. The window is per groupBy value; delivery stays one message listing newly eligible groups. Either CALENDAR (kind: CALENDAR, calendarUnit: WEEK | MONTH) = at most once per current ISO week / calendar month, or ROLLING (kind: ROLLING, windowDays: N) = at most once every N days. Optional here: when provided the result also reports how many notifications would actually be sent after dedup.
contextYesA few words on why you are calling this tool, for the server developer's analytics. Describe people and things by role, such as "a customer". Do not quote the user or include names, emails, or other personal details.
queriesYesSame series objects as the `query` tool `queries` array (referenced by `condition`).
conditionYesAlerts v3 firing rule: a single boolean expression over the query names (`name` field of each query). Supports arithmetic (+ - * /), comparisons (> >= < <= == !=), logical and/or/not, parentheses, and these window functions: rollingSum(a, N, UNIT) (trailing sum over the last N units, UNIT ∈ DAY|WEEK|MONTH, inclusive of today), weekToDateSum(a) (Monday-to-date), monthToDateSum(a) (1st-of-month-to-date), and timeShift(a, N, UNIT) (value shifted back N units; may wrap a window function). Examples: `a > 1000`, `rollingSum(a, 7, DAY) > 1000`, `(a - timeShift(a, 1, DAY)) / timeShift(a, 1, DAY) > 0.2`, `a > 10000 or rollingSum(a, 7, DAY) > 50000`.
llm_modelNoThe model identifier you are running as, for example "claude-opus-4-8" or "gpt-5.2". Pass "unknown" if you are not certain. Recorded for the server developer's analytics.
lookbackDaysNoTrailing window length in days to replay the condition over (default 45).
conversation_idNoThe conversation_id this server returned earlier in this conversation, unchanged, including after later user messages. Omit it if you have none. Recorded for the server developer's analytics.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed2 schema fields changed
    • changedInput schema / properties / context / description
      Previous value: -"Explain in 15-25 words, in third person, why this tool is called and how it supports the user's goal. For analytics only. You MUST describe only the abstract purpose of the tool call. NEVER include, repeat, paraphrase, or infer personal, sensitive, or identifying information from the user request or tool results, including names, emails, phone numbers, IPs, IDs, or credentials. You MUST generalize specific entities into roles such as \"a user\", \"the customer\", or \"an account\". Example: \"Retrieving a customer's recent orders to investigate a billing issue and help support determine the appropriate resolution.\""New value: +"A few words on why you are calling this tool, for the server developer's analytics. Describe people and things by role, such as \"a customer\". Do not quote the user or include names, emails, or other personal details."
    • changedInput schema / required
      Previous value: -[
-  "queries",
-  "condition",
-  "context",
-  "llm_model"
-]New value: +[
+  "queries",
+  "condition",
+  "context"
+]
  2. Changed4 schema fields changed
    • addedInput schema / properties / context
      Added value: +{
+  "description": "Explain in 15-25 words, in third person, why this tool is called and how it supports the user's goal. For analytics only. You MUST describe only the abstract purpose of the tool call. NEVER include, repeat, paraphrase, or infer personal, sensitive, or identifying information from the user request or tool results, including names, emails, phone numbers, IPs, IDs, or credentials. You MUST generalize specific entities into roles such as \"a user\", \"the customer\", or \"an account\". Example: \"Retrieving a customer's recent orders to investigate a billing issue and help support determine the appropriate resolution.\"",
+  "type": "string"
+}
    • addedInput schema / properties / conversation_id
      Added value: +{
+  "description": "The conversation_id this server returned earlier in this conversation, unchanged, including after later user messages. Omit it if you have none. Recorded for the server developer's analytics.",
+  "type": "string"
+}
    • addedInput schema / properties / llm_model
      Added value: +{
+  "description": "The model identifier you are running as, for example \"claude-opus-4-8\" or \"gpt-5.2\". Pass \"unknown\" if you are not certain. Recorded for the server developer's analytics.",
+  "type": "string"
+}
    • changedInput schema / required
      Previous value: -[
-  "queries",
-  "condition"
-]New value: +[
+  "queries",
+  "condition",
+  "context",
+  "llm_model"
+]
  3. Changed1 schema field changed
    • changedInput schema / required
      Previous value: -[
-  "condition"
-]New value: +[
+  "queries",
+  "condition"
+]
  4. Changed1 schema field changed
    • changedInput schema / properties / dedup / description
      Previous value: -"Deduplication config controlling how often a still-firing alert notifies. Either CALENDAR (kind: CALENDAR, calendarUnit: WEEK | MONTH) = at most once per current ISO week / calendar month, or ROLLING (kind: ROLLING, windowDays: N) = at most once every N days. Optional here: when provided the result also reports how many notifications would actually be sent after dedup."New value: +"Deduplication config controlling how often a still-firing group notifies. The window is per groupBy value; delivery stays one message listing newly eligible groups. Either CALENDAR (kind: CALENDAR, calendarUnit: WEEK | MONTH) = at most once per current ISO week / calendar month, or ROLLING (kind: ROLLING, windowDays: N) = at most once every N days. Optional here: when provided the result also reports how many notifications would actually be sent after dedup."
  5. First observed

TDQS

A4.8/5.0
Behavior5/5

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

The annotations already declare this is a safe read-only operation, but the description adds substantive context beyond them: nothing is created, delivery is simulated, the dedup window is scoped per groupBy value, and the response reports firingDays, firingRows, notificationsCount, and a sample of firing dates with per-group state. For a tool with no output schema, this return-value disclosure is exactly what the description must carry.

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

Conciseness4/5

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

The opening sentence front-loads purpose and sequencing before any detail, and the return-value list is compact. The inline JSON example is long and duplicates syntax already present in the condition schema, which is the main thing keeping it from a 5.

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 an 8-parameter tool with a custom expression language and no output schema, the description covers the evaluation window, the full return shape, dedup semantics, and the relationship to create_alert. Nothing an agent needs in order to call it correctly or interpret its result 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?

With 100% schema coverage, the baseline is 3, but the description adds meaning beyond the schema: the queries/condition/dedup triple mirrors create_alert minus the notification channel, lookbackDays defaults to 45 and is a trailing replay window, and dedup affects notification counts rather than just filtering. It doesn't independently document slug, context, llm_model, or conversation_id, but those are fully described in 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?

The description opens with a specific verb and resource: "Backtest a cost-alert condition BEFORE creating it", and immediately contrasts the tool with the sibling that actually persists state (create_alert). It states that it replays the condition over a lookback window and reports fire counts, so an agent knows exactly what the tool produces without opening the 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?

It gives explicit when-to-use guidance ("Use this to sanity-check a condition/threshold (and tune dedup) before calling create_alert") and clarifies the boundary with create_alert by noting no notification channel is needed. The example question at the end reinforces the intended scenario with a concrete trigger.

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

All Costory: Your Finops MCP toolsInspect
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources