AI for FinOps — Cloud Cost Optimization with MCP

Open-source Model Context Protocol (MCP) resources for Cloud FinOps: cloud cost optimization, AI cost management, FinOps automation, tagging governance, cost anomaly detection, and budget monitoring — across AWS, Azure, and GCP.

Links License Good First Issues

What is this repository?

A practical resource hub for FinOps practitioners, cloud engineers, and platform teams who want to use AI agents to automate cloud cost optimization. This repo provides tutorials, MCP server documentation, client guides, and security frameworks for implementing the Model Context Protocol across AWS, Azure, and GCP.

19 MCP servers documented | 7 step-by-step tutorials | 9 MCP clients compared | 3 cloud providers covered

How do I get started?

  1. Choose an MCP client — Claude, ChatGPT, Gemini, Copilot, Cursor, Kiro, VS Code
  2. Run your first MCP tutorial — query real-time AWS pricing in 15 minutes
  3. Explore all FinOps MCP servers — AWS, Azure, GCP, Tagging, JIRA, Slack

Recent Updates (July 2026)

  • Azure FinOps MCP Server (public preview) (June 23, 2026) — Microsoft's first official FinOps-branded MCP server, the cost-and-usage packaging of the ARM MCP Server (preview since May 7). Natural-language Azure Resource Graph queries for estate-wide waste detection; the official @azure/mcp server also gained a retail pricing tool for pre-deployment cost estimation. Updated: server doc, tutorial. Source: Azure Blog.
  • AWS MCP Server GA (May 6, 2026) — Now part of the broader Agent Toolkit for AWS. The previous aws-mcp:InvokeMcp permission is replaced by standard IAM policies using the aws:ViaAWSMCPService and aws:CalledViaAWSMCP context keys. New run_script tool runs short Python scripts in a server-side sandbox for multi-step workflows. Frankfurt (eu-central-1) endpoint added alongside us-east-1. Updated: tutorial, server doc.
  • Google managed MCP servers GA (April 29, 2026): more than 50 official Google MCP servers across infrastructure, databases, analytics, storage, and Workspace APIs (BigQuery, Cloud Run, Apigee, AlloyDB, Spanner, Firestore, GKE). Updated: server doc.
  • New: Workflow & Collaboration Servers — FinOps Tagging Compliance, JIRA, and Slack MCP servers for automated cost governance
  • Registry Backfill — registry.yaml machine-readable registry, now at 18 entries with the ARM MCP Server addition
  • MCP Authentication Vulnerabilities (refreshed May 2026): critical security risks plus 2 new 2026 CVEs (CVE-2026-0621 in the TypeScript SDK, CVE-2026-40933 in Flowise) (read more)
  • MCP Dev Summit North America 2026 (NYC): AAIF reported 1,200 attendees, 110+ million monthly SDK downloads, and 170 member organizations. Headline disclosures: a DNS rebinding attack class against MCP servers (Jonathan Leitschuh / Braise) and Amazon's "lethal trifecta" pattern for scanning agent configurations. Source: AAIF blog

What's inside?

Section What you'll find
/foundations What is MCP, how it works, architecture deep-dive
/servers 19 MCP servers — AWS, Azure, GCP, Tagging, JIRA, Slack
/clients 9 MCP clients compared — Claude, ChatGPT, Gemini, Copilot, Cursor, Kiro
/tutorials 7 step-by-step guides for AWS, Azure, GCP cost analysis
/governance Security best practices, IAM policies, vulnerability guides

What is MCP and why does it matter for FinOps?

MCP connects AI clients to cloud cost management tools through a standardized protocol

The Model Context Protocol (MCP) is an open standard that lets AI agents securely connect to external tools — like AWS Cost Explorer, GCP BigQuery billing exports, Azure Cost Management, or third-party FinOps platforms like Vantage. It's the bridge between "ask a question about cloud spend" and "get a real answer from live data."

Industry Adoption (May 2026): MCP was donated to the Linux Foundation (December 2025) with backing from Anthropic, OpenAI, Google, Microsoft, and AWS. At MCP Dev Summit North America 2026 in New York City, the Agentic AI Foundation reported 1,200 attendees (doubled from the previous summit), 110+ million monthly SDK downloads, and 170 member organizations in under 4 months. All major AI platforms now support MCP.

What can MCP do for cloud cost optimization?

  • Cost anomaly detection — AI agents query Cost Explorer and alert on spend spikes
  • Tagging compliance — automated audits of tag coverage and drift detection
  • Resource optimization — rightsizing recommendations, idle resource identification
  • Budget monitoring — real-time threshold alerts via Slack, JIRA ticket creation
  • Multi-cloud cost analysis — unified queries across AWS, Azure, and GCP
  • Cost simulation — what-if scenarios for commitment purchases and architecture changes

But MCP also raises governance and security challenges — this repo addresses both.

Contributing

We welcome contributions — new MCP servers, tutorials, use cases, and security guidance.

  1. Fork the repo and create a branch
  2. Add your content
  3. Open a PR with a clear description

See CONTRIBUTING.md for details. New to MCP or FinOps? Start with issues labeled good first issue.

Community