This project exposes AWS cost data to MCP clients. One half reads AWS Cost Explorer to answer questions about spend broken down by service, account and period. The other half reads Amazon Bedrock model-invocation logs from CloudWatch Logs, which turns opaque Bedrock charges into per-model usage figures — how many tokens each model consumed and what that cost. Both are Python tools over the AWS SDK, so the credentials and IAM permissions of the server determine what is visible; the README notes that if the server's IAM role can assume roles in other accounts, it can report their spend too, making a single deployment usable for a whole organisation. Two deployment shapes are documented. Locally, the server runs over stdio and is registered in Claude Desktop. Remotely, it runs on an EC2 instance and speaks MCP over HTTPS with SSE, with a section on securing that endpoint; the repository also ships an MCP client built into a LangGraph agent for that path. Diagrams in the README trace the request flow from client through the server to the Cost Explorer and CloudWatch APIs. The audience is cloud and FinOps engineers who want to ask about AWS and Bedrock spend in natural language instead of building Cost Explorer queries by hand.
https://github.com/aarora79/aws-cost-explorer-mcp-server
Apache APISIX is a high-performance cloud-native API Gateway and AI Gateway that supports Model Context Protocol (MCP) integration through its mcp-bridge plugin, offering dynamic traffic management, security, and observability features.
AWS MCP Servers is a suite of specialized Model Context Protocol servers that integrate AWS best practices and resources into AI-powered development workflows to enhance cloud-native development and infrastructure management.
Webiny is an open-source, self-hosted headless CMS and framework, built on AWS serverless with AI-assisted development capabilities via an embedded MCP server, designed for large organizations.
MCP Registry is a community-driven centralized service offering a RESTful API for discovering, managing, and publishing Model Context Protocol (MCP) servers with MongoDB and Docker support.
Olares is an open-source personal cloud operating system that enables users to run local AI models and applications with personalized, context-aware AI interactions using a Model Context Protocol-like mechanism, providing a private and secure AI home server environment.
Cloudflare MCP Server is a suite of MCP servers that enable large language model clients to interact with and manage Cloudflare's cloud services through natural language commands, facilitating automation and insights across various domains.
mcp-grafana is an MCP server that provides comprehensive programmatic access to Grafana's dashboards, datasources, alerts, incidents, and on-call management features through MCP clients.
MCP Containers offers containerized versions of hundreds of MCP servers, enabling easy, secure, and up-to-date deployment of Model Context Protocol servers for AI integration and experimentation.