This MCP server, FastMCP, acts as a multi-cloud FinOps assistant. It connects Gemini-powered assistants with cost insights across AWS, GCP, and Azure, enabling natural language queries to generate cost breakdowns, run FinOps audits, and produce usage summaries all from your local environment. The server supports a CLI- and FastAPI-compatible architecture, allowing integration with MCP clients and automation pipelines for budgeting, reporting, and optimization prompts. You can ask questions like how much was spent on a given cloud, request a cross-cloud audit, or obtain a consolidated budget status across all providers. To use it, start the MCP server and connect your MCP client (e.g., Claude Desktop, Amazon Q, or any compatible MCP client) to pose prompts and receive structured responses with actionable recommendations.

Prerequisites:

• Python 3.11+ installed on your machine
• Poetry for dependency management
• Access to AWS CLI, Google Cloud SDK (gcloud), and Azure CLI (az) if you intend to use real credentials

Installation steps:

1. Clone the repository git clone https://github.com/Eazy-Ops/multi-cloud-finops-mcp-server.git cd multi-cloud-finops-mcp-server

2. Install Python dependencies with Poetry poetry install

3. Activate the Poetry-managed virtual environment poetry shell

4. Run the MCP server (examples)

   Option A: using the MCP entry point

   poetry run python -m mcp.server.fastmcp

   Option B: direct Python invocation (if preferred)

   python -m mcp.server.fastmcp

Note: If you prefer running the server via CLI prompt, you can also start the main module directly if you have a suitable entry point configured via your runtime environment.

Environment and credentials:

• The server uses local SDKs/CLIs for credentials; ensure AWS, GCP, and Azure CLIs are installed and configured if you plan to query real data.
• AWS: run aws configure --profile <profile>
• GCP: gcloud auth application-default login or provide a service_account_key_path in your calls
• Azure: az login; for service principals, set AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET

Tips:

• Keep your credentials local; the MCP server aims to avoid leaking credentials outside your machine.
• Use the provided CLI prompts and natural-language prompts to generate cost analyses, audits, and budget summaries across providers.
• If you add new cloud accounts, ensure corresponding CLI tools are authenticated and configured before issuing queries.