The Agent Toolkit for AWS is an official AWS-supported collection of Model Context Protocol servers, skills, and plugins designed to enable AI coding agents to build, deploy, and manage applications on AWS. Written primarily in Python, the toolkit provides a unified interface that works with multiple popular coding agents including Claude Code, Codex, Cursor, and Kiro, allowing developers to leverage AI agents within their existing workflows.
The toolkit is organized into four main plugin offerings. The aws-core plugin serves as the foundational component, covering service selection, infrastructure-as-code with CDK and CloudFormation, serverless computing, container orchestration, storage solutions, observability, billing, SDK usage, and deployment workflows. The aws-agents plugin enables building AI agents on AWS using Amazon Bedrock and AgentCore. The aws-data-analytics plugin focuses on data lake construction, analytics operations, and ETL workflows using S3 Tables, AWS Glue, and Athena. The aws-agents-for-devsecops plugin integrates with AWS DevOps Agent and AWS Security Agent to support incident investigation, code review, UAT execution, vulnerability scanning, and penetration testing.
At the core of the toolkit is the AWS MCP Server, a managed service providing full coverage of over 300 AWS services through a single authenticated endpoint. The server includes sandboxed Python script execution for complex multi-step operations, real-time AWS documentation access without requiring authentication, and enterprise-grade controls including CloudWatch metrics, IAM context keys for agent-specific policies, and CloudTrail audit logging. This architecture allows agents to discover and load skills on demand, retrieving only what is relevant to the current task.
The repository demonstrates active maintenance and responsiveness to issues. The repository's contributor network overlaps with major projects including Deno, TypeScript, and Visual Studio Code, suggesting cross-pollination with the broader developer tooling ecosystem.
Installation varies by agent platform. Claude Code users access plugins through the official Anthropic marketplace. Cursor users import the repository as a team marketplace through plugin settings. Kiro users configure the AWS MCP Server in their MCP configuration file and install skills using the uv package manager. The toolkit emphasizes security best practices, recommending that users pin to specific versions on PyPI to ensure reproducible behavior and protect against supply chain risks.
The Agent Toolkit for AWS represents AWS's official successor to earlier MCP servers and skills released through AWS Labs. The toolkit distinguishes itself through IAM condition keys that differentiate agent actions from human actions, enabling fine-grained access control policies, comprehensive CloudWatch metrics and CloudTrail logging for monitoring and auditing agent activity, and agent skills that have undergone thorough end-to-end evaluation. The project is licensed under Apache-2.0 and includes comprehensive documentation covering setup, configuration, and reference materials for the AWS MCP Server tools.
