The AWS Infrastructure as Code MCP Server: AI-Powered CDK (aws.amazon.com)

AWS announced the Infrastructure-as-Code (IaC) MCP Server, a local service that connects AI assistants (Kiro CLI, Claude, Cursor, etc.) to AWS CloudFormation and CDK workflows via the open Model Context Protocol (MCP). The server provides contextual, natural-language access to documentation, code samples, and automated checks so assistants can search CDK/CloudFormation docs, suggest constructs, validate templates, and troubleshoot failed deployments—accelerating day-to-day IaC development and reducing manual lookups and guesswork. Technically, the IaC MCP Server exposes nine tools split between remote documentation search (e.g., search_cdk_documentation, search_cdk_samples_and_constructs, read_cdk_documentation_page) and local validation/troubleshooting (cdk_best_practices, validate_cloudformation_template using cfn-lint, check_cloudformation_template_compliance with cfn-guard, troubleshoot_cloudformation_deployment with CloudTrail analysis, and pre-deploy validation guidance). The server runs locally (installed via uv), communicates with AI clients over stdio (no open network ports), and uses your AWS credentials; validation is local and needs no AWS perms, while troubleshooting requires read-only CloudFormation and CloudTrail permissions (DescribeStacks/Events/Resources, LookupEvents). Privacy caveat: API responses may be shared with third‑party AI providers, so organizations must vet data handling. The project is available now on GitHub and promises faster, safer IaC iterations—provided teams manage IAM scope and provider trust.