# Agent Toolkit for AWS

> Official AWS-supported MCP server, skills, and plugins that help AI coding agents build, deploy, and manage applications on AWS.

Agent Toolkit for AWS is an open-source repository from AWS that packages an AWS MCP Server configuration, agent skills, plugins, and rules files so AI coding agents can work with AWS services. The README lists support for Claude Code, Codex, Cursor, Devin, Kiro, and fx, and marks the project status as GA.

## What It Is

The toolkit gives coding agents tools, knowledge, and guardrails for AWS work. Plugins bundle the AWS MCP Server configuration with agent skills in a single install, while skills are curated instruction and reference packages that agents load on demand for specific AWS tasks. Rules files are recommended project-level configurations that tell agents to use the MCP server, discover skills, and search documentation before acting.

## Plugins and Skills

The repository lists several plugins:

- aws-core covers service selection, CDK/CloudFormation, serverless, containers, storage, observability, billing, SDK usage, and deployment.
- aws-agents covers building AI agents with Amazon Bedrock and AgentCore.
- aws-data-analytics covers data lake, analytics, and ETL workflows with S3 Tables, AWS Glue, and Athena.
- aws-agents-for-devsecops covers incident investigation, code review, UAT, vulnerability scanning, and penetration tests using AWS DevOps Agent and AWS Security Agent.
- aws-startup-advisor covers startup-focused architecture, cost, and security guidance, plus migrations from other clouds and AI stacks.

Plugins are available for Claude Code, Codex, Cursor, and Devin. Other agents can configure the MCP server directly and install skills with `npx skills add`.

## AWS MCP Server

The AWS MCP Server is a managed server that exposes AWS through the Model Context Protocol. Per the README, it offers coverage of 300+ AWS services through one authenticated endpoint, sandboxed Python script execution, and real-time documentation search that needs no authentication. Enterprise controls include CloudWatch metrics, IAM context keys for agent-specific policies, and CloudTrail audit logging.

## Relationship to AWS Labs

The README describes the toolkit as the successor to the MCP servers, skills, and plugins released under AWS Labs starting in 2025. AWS Labs tools continue to work and accept contributions, and AWS says the best of them will move into this toolkit over time.

## Setup Path

Setup varies by agent. The AWS CLI route uses `aws configure agent-toolkit`. Kiro and fx setups combine an MCP server entry run through uvx with locally installed skills. These require uv, and an AWS account with local credentials is needed for API calls and script execution but not for documentation search or skill discovery.

## Features
- AWS MCP Server with access to 300+ AWS services through a single authenticated endpoint
- Sandboxed Python script execution
- Real-time AWS documentation search
- Agent skills loaded on demand
- Plugins bundling MCP configuration and skills
- Rules files for agent configuration
- IAM condition keys distinguishing agent from human actions
- CloudWatch metrics and CloudTrail audit logging
- AWS CLI integration via aws configure agent-toolkit

## Integrations
Claude Code, Codex, Cursor, Devin, Kiro, fx, AWS CLI, Amazon Bedrock, AWS Glue, Amazon Athena, AWS DevOps Agent, AWS Security Agent, AWS CloudTrail, Amazon CloudWatch, AWS IAM

## Platforms
WEB, API, DEVELOPER_SDK, CLI

## Pricing
Open Source

## Links
- Website: https://github.com/aws/agent-toolkit-for-aws
- Documentation: https://docs.aws.amazon.com/agent-toolkit/latest/userguide/
- Repository: https://github.com/aws/agent-toolkit-for-aws
- EveryDev.ai: https://www.everydev.ai/tools/agent-toolkit-for-aws