# Costory FinOps MCP [Health: Active]

**Category:** 🗄️ Databases  
**Repository:** https://github.com/costory-io/costory-finops-mcp-skills  
**GitHub Stars:** 2  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/costory-finops-mcp

## Description
FinOps MCP: query allocated, correlated cloud and AI cost across AWS, GCP, Azure and Snowflake.

## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: high). Install path detected from listing signals. Add as a URL/SSE server in your client:

```json
"mcpServers": {
  "costory-finops-mcp": {
    "url": "https://app-api.costory.io/mcp"
  }
}
```

## Documentation

## What the Costory FinOps MCP MCP server does

The Costory FinOps MCP MCP server gives supported AI clients access to Costory workspace data through a hosted Model Context Protocol endpoint. Costory normalizes billing from AWS, GCP, Azure, Snowflake, Datadog, OpenAI, and Anthropic into a common schema. It can also allocate shared or untagged spend using usage metrics and associate cost changes with deployments, incidents, alerts, and other events.

The repository contains the workflow skills layered over the MCP tools. These skills guide an assistant through investigations such as explaining a production cost increase, comparing budgets with actuals, analyzing BigQuery warehouse spend, or designing a custom cost split.

## How it works

The remote endpoint is `https://app-api.costory.io/mcp`. Clients authenticate with OAuth in a browser, and requests are scoped to the calling user's Costory workspace role. The service does not require IAM credentials, Docker, or a locally running server.

The workflow generally starts with orientation tools that discover available dimensions, dashboards, metrics, and suggested query approaches. Query tools then retrieve cost, usage, metric, formula, and budget data, including period-over-period comparisons. Skills define sequencing for recurring tasks instead of requiring the assistant to determine every tool call independently.

Write operations remain inside the connected Costory workspace. Virtual dimensions use ordered CEL rules and can be drafted, updated, previewed, and published. Reports, dashboards, alerts, and events can also be created or changed through the server.

## Setup and configuration

A Costory workspace with billing data connected is required. Add the hosted endpoint as a custom HTTP MCP connector and complete the OAuth flow. The repository includes an `.mcp.json` example with an OAuth callback port of `8080`.

Clients without native remote MCP support can use `mcp-remote` through an `npx` command. The repository also provides plugin packaging for Claude Code, Codex, and Cursor. Claude Code and Codex can install the plugin from the Costory marketplace commands documented in the repository.

## Tools and capabilities

The Costory FinOps MCP MCP server groups its tools into five areas:

- Orientation: discover context, search records, inspect objects, and suggest dimensions, usage metrics, or actions.
- Query: retrieve costs, usage, metrics, formulas, and budgets.
- Allocation: define and test custom cost axes with virtual-dimension rules.
- Reporting: build dashboards and schedule Slack, Teams, or email reports.
- Alerting and events: create or review cost and budget alerts and annotate events for correlation.

Available skills include query investigation, cost-change analysis, virtual dimensions, dashboards, reports, BigQuery-specific guidance, and recipes. Recipes cover scenarios such as GCP committed-use discounts, EC2 spike alerts, untagged-cost coverage, provider credits, namespace cost, and compute drilldowns.

## Limitations and notes

Tool names and payloads are versioned, so the Costory API reference is the authoritative source for current request details. Explorer period-over-period queries do not handle every “what changed” workflow; the repository’s query skill hands those investigations to the reports-based explanation flow. Access and write scope follow the user's Costory workspace role, and the repository does not provide a standalone local server.

_Full upstream README: https://allmcps.com/mcp/costory-finops-mcp/readme_