AWS released an open-source Agent Toolkit that gives AI coding assistants access to current AWS documentation, APIs, and best practices. The toolkit packages curated instructions as skills, plugins, and rules that guide agents through specific cloud tasks without relying on outdated model knowledge.

The toolkit includes three main components: skills that walk agents through operations like creating S3 Tables lakehouses or debugging Lambda timeouts, plugins that group related capabilities, and rules that enforce infrastructure-as-code preferences. An MCP server integration provides live access to AWS documentation and native auditing controls.

Most coding agents suffer from knowledge cut-off issues—they're trained on data months old, causing problems when AWS services evolve rapidly. A generic agent might incorrectly use LOCATION clauses with S3 Tables, not knowing the service manages storage differently than external tables.

The toolkit prevents these mistakes by checking existing resources, using correct APIs, and verifying assumptions against current documentation. In one demonstration, an AI agent successfully built a complete data pipeline with private JDBC source, VPC networking, S3 Tables, Glue ingestion, and Athena validation in just over 30 minutes—work that would typically require deep AWS expertise.