Costory
Hỏi trợ lý AI của bạn một câu hỏi về chi phí. Nhận phân bổ, tương quan và giải thích trong một phản hồi. Costory kết nối Claude, Codex hoặc Cursor với dữ liệu chi phí chuẩn hóa trên AWS, GCP, Azure, Datadog, OpenAI và Anthropic.
Tài liệu
Costory: The FinOps & Cloud Cost Automation MCP
The official FinOps MCP (Model Context Protocol) and Cloud Cost Automation MCP by Costory.
Connect Claude, Cursor, or any MCP-compatible AI assistant directly to your AWS, Azure, or GCP billing data. The Costory MCP provides normalized, allocated cost data correlated with infrastructure events, turning your LLM into an autonomous FinOps engineer that understands how deployments, traffic spikes, and pricing shifts actually impact your cloud bill.
Why Costory MCP?
Feeding raw AWS or GCP billing lines into a prompt doesn't work. The Costory FinOps MCP acts as a context layer, delivering normalized cost data that is already allocated and correlated with your infrastructure events. It turns your AI assistant into a FinOps practitioner that understands how deploys, traffic, and pricing shifts actually impact your bill.
Installation
Install the Costory FinOps MCP plugin via the Claude Code marketplace:
/plugin marketplace add costory-io/costory-plugin
/plugin install costory@costory-marketplace
Setup
After installing, run /mcp in Claude Code and follow the browser OAuth flow to log in with your Costory account. This connects your local automation MCP environment to your cloud infrastructure.
FinOps Skills
Easily trigger built-in FinOps automation skills:
| Skill | Description |
|---|---|
/costory:analyze-costs | Analyze cloud costs by service, team, environment, or any dimension |
/costory:investigate-change | Investigate cost spikes and compare periods |
/costory:query-metric | Query custom business metrics (DAU, requests/day, etc.) |
/costory:setup-alert | Create cost alerts with Slack notifications |
/costory:send-report | Send cost reports to Slack channels |
Autonomous FinOps Agents
| Agent | Description |
|---|---|
cost-investigator | Autonomous deep-dive Automation MCP agent that drills down across dimensions to find root causes of cost changes |
Real-World Examples
Here is what it looks like when you bring AWS and GCP costs directly into your AI assistant to automate FinOps workflows:
1. Root Cause Analysis for AWS Cost Spikes (Month-over-Month)
Stop logging into billing consoles. Ask your MCP agent to query exactly what you need to investigate an anomaly:
"How did our AWS costs change vs last month?"
MCP Execution (query payload):
{
"queries": [{ "type": "cost", "name": "a", "metricId": "cost", "currency": "USD" }],
"from": "2026-03-01",
"to": "2026-03-31",
"compare": { "from": "2026-02-01", "to": "2026-02-28" }
}
### 2. Automation FinOps: Advanced Smart Alerts
Set up complex anomaly detection natively within your chat using CEL expressions:
> *"Create a Slack alert if our 7-day rolling AWS cost exceeds $1,000 AND has spiked by more than $1,000 compared to the previous week."*
**How the MCP handles it (`create_alert`):**
```json
{
"name": "7-Day Cost Spike Alert",
"queries": [{ "type": "cost", "name": "a", "metricId": "cost", "currency": "USD", "filterCel": "cos_provider in [\"AWS\"]" }],
"condition": "rollingSum(a, 7, DAY) > 1000 && (rollingSum(a, 7, DAY) - timeShift(rollingSum(a, 7, DAY), 7, DAY)) > 1000",
"dedup": { "kind": "CALENDAR", "calendarUnit": "WEEK" },
"notificationChannel": "SLACK",
"slackChannelId": "C01ABC"
}
3. Automated Cost Reports via Slack
Put your FinOps reporting on autopilot with scheduled Top/Flop movers:
"Send a weekly Slack digest showing our top 10 and bottom 5 cost movers by service, comparing this week to last week."
How the MCP handles it (create_report):
{
"scheduledPeriod": "WEEKLY",
"widget": {
"type": "TOP_FLOP",
"queries": [{ "type": "cost", "name": "a", "metricId": "cost", "currency": "USD", "groupBy": "cos_service_name" }],
"topN": 10,
"flopN": 5
},
"destinations": [{ "destinationType": "SLACK", "channelId": "C01ABC" }]
}
MCP Tools Reference
This plugin connects to the Costory MCP server, exposing a suite of FinOps and Automation MCP tools organized by workflow:
Search and contextualize your infrastructure
get_context: Fetch the active operational contextlist_organizations: List accessible organizationssearch: Discover dimension values, events, alerts, dashboards, and virtual dimensions. Example: A developer asks about "Project Phoenix". Use search to instantly find all associated AWS tags, dashboards, and historical cost events for that specific project.
Query cost, usage and external metrics into a unified endpoint
query: Core data tool for running unified queries for cloud costs, business metrics, usage metrics, budgets, and formulassuggest_groupby: Find the most impactful dimension to split costs by. Example: You find a cost spike but have no idea what labels or tags are available to drill down. Usesuggest_groupbyto automatically find the correct axes (like team, service, or environment) to guide your investigation.list_metrics: List available business metric datasourcessuggest_usage_metrics: Suggest infrastructure usage metric units relevant to a billing scopeget: Fetch full resource data by ID (dashboards, budgets, or cost alerts)
Generate Dashboards and Visualization
get_dashboard_widget_data: Run a saved dashboard widget and return its dataget_dashboard_widget_image: Retrieve a visual image snapshot of a widgetcreate_dashboard: Create a new dashboard with specific widgets and contextupdate_dashboard: Append, replace, or remove widgets on an existing dashboardset_dashboard_tags: Assign descriptive tags to dashboardsset_dashboard_team: Assign a dashboard to a specific team
Automation & Smart Alerts
list_alerts: View and manage existing active cost and budget alertspreview_alert: Test alerting conditions on historical datacreate_alert: Set up FinOps automation alerts for spending thresholds. Example: You notice a database's cost slowly creeping up over time. Set a smart alert: "Warn me when the 7-day rolling cost for this database exceeds $1,000 so I can investigate before it gets out of hand."
Event correlation
list_events: Correlate cost changes with tracked infrastructure eventscreate_event: Annotate cost changes and attach them to a context. Example: You detect a large spike and learn from the data team it was a one-time mislabeled backfill. Attach an event to the cost to remember what happened and add this context to your FinOps memory.
Automated Cost Reports & Slack Notifications
list_available_destinations: Discover available Slack channels, Teams channels, and email destinationscreate_report: Autonomously generate and schedule rich cost reports. Example: You are working on deprecating a legacy Terraform module. Create a weekly report that sends a chart showing the top cost decreases directly to your Slack so you can track the savings progress.
Organization Metadata
list_teams: List teams the current user belongs tolist_tags: View all available organization tagsdelete_tag: Remove a specific tag
Documentation
search_documentation: Query the internal FinOps knowledge baseget_documentation_page: Fetch the full content of a documentation page
Agent Helpers
suggest_actions: Receive context-aware follow-up FinOps suggestions based on current investigation context