Practical Model Context Protocol (MCP) resources for Cloud FinOps: pricing, budgets, anomaly checks, automation — with security guardrails.
This repository is up-to-date as of January 2026 with the latest MCP developments:
- Tutorial Updates (December 2025) - Major improvements across all quickstart tutorials:
- GCP BigQuery Tutorial - Added local vs remote MCP comparison, fixed video embedding, improved setup instructions
- AWS Kiro CLI Tutorial - Renamed from "Amazon Q" for clarity, enhanced installation steps
- Azure MCP Tutorial - Added critical clarification that official Azure MCP lacks cost/billing APIs, comprehensive comparison of community alternatives with production-readiness assessment, and Cline-specific configuration guidance
- AWS MCP Server Unified Architecture - New consolidated server announced November 2025 with access to 15,000+ AWS APIs, Agent SOPs, and AWS's first remote MCP server (learn more)
- MCP Specification 2025-11-25 - Task workflows, enhanced OAuth PKCE, client credentials, and cross-app authorization (architecture guide)
- Linux Foundation Donation - MCP donated to Agentic AI Foundation (December 2025) with support from Anthropic, OpenAI, Google, Microsoft, AWS, and others (details)
- New Client Documentation - Added guides for ChatGPT, Gemini, Copilot, Claude Code, and Kiro - now covering all 9 major MCP clients (view all clients)
- Enhanced Security - Updated security best practices for MCP 2025-11-25 spec with remote MCP deployment guidance (security docs)
- Improved Navigation - Reorganized repository structure with INDEX files for better discoverability (see CHANGELOG)
- ✅ Download a client
- ✅ Run your first MCP
- ✅ Explore the FinOps MCPs and pick the one that best fits your use case
- /foundations → background notes, whitepapers, blog summaries
- /servers → registry of available MCP servers (pricing, tagging, governance)
- /clients → tested MCP clients (Claude, Cursor, VS Code, Q, etc.)
- /tutorials → runnable guides (step by step)
- /use-cases → applied scenarios (budgeting, anomaly detection, tagging compliance)
- /governance → security checklists, threat models, deployment guidance
- /presentations → slides, abstracts, LinkedIn drafts
- /resources → external links (FinOps WG docs, repos, talks, videos)
MCP connects AI clients to multiple data sources and services through a standardized protocol
MCP is an open standard protocol that lets LLMs act as agents by safely connecting to external tools (servers) like AWS Cost Explorer, a GCP BigQuery dataset with billing exports, an Azure storage account holding cost data, or 3rd-party cloud finops solutions like Vantage.
Industry Adoption (January 2026): In December 2025, Anthropic donated MCP to the Agentic AI Foundation (Linux Foundation), with founding support from Anthropic, Block, and OpenAI, plus backing from Google, Microsoft, AWS, Cloudflare, and Bloomberg. This ensures MCP remains open, neutral, and community-driven as critical AI infrastructure.
As of January 2026, the ecosystem has grown to 10,000+ active MCP servers and is now supported by major AI platforms including ChatGPT, Claude, Gemini, Microsoft Copilot, VS Code, Cursor, Kiro, and Amazon Q.
In FinOps, MCP unlocks:
- Faster cost simulations
- Real-time tagging compliance
- Forecasting and Cost Simulations
- Cost Optimization recommendations
But also raises governance and security challenges — this repo addresses both sides.
We welcome contributions:
- Fork the repo and create a branch
- Add your MCP server, tutorial, or use case
- Open a PR with a clear description
See CONTRIBUTING.md for details.
New to MCP or FinOps? Start with issues labeled good first issue.
- Join discussions in the FinOps Foundation Slack
- Follow updates on LinkedIn
- Share your use cases, raise issues, propose servers
This project is licensed under the Apache 2.0 License.