What Is the AWS Agent Toolkit for AWS? A Complete Guide to AI Agent Integration

The AWS Agent Toolkit for AWS is a curated collection of plugins, skills, and rules that enable AI coding agents like Claude Code, Codex, and Cursor to interact with AWS services securely through a managed MCP server.

The AWS Agent Toolkit for AWS (hosted at aws/agent-toolkit-for-aws) provides a structured framework for integrating large language model agents with Amazon Web Services. This open-source repository packages domain-specific expertise into reusable components that agents can discover and execute, ensuring operations remain compliant with enterprise security policies.

Core Architecture of the AWS Agent Toolkit

The toolkit implements a three-layer architecture that separates authentication, domain logic, and policy enforcement.

MCP Server (Model Context Protocol)

At the foundation lies the MCP Server, a managed Model Context Protocol endpoint that handles authentication and API mediation. According to the source code in the repository root, this server provides full AWS API coverage, executes sandboxed Python scripts, and serves real-time AWS documentation to connected agents. The MCP server acts as a secure gateway, ensuring every request is authorized and logged before reaching AWS infrastructure.

Domain-Specific Plugins

Plugins package related skills and configurations for specific AWS domains. Each plugin resides in its own directory under plugins/ and contains an MCP server configuration along with instructional content. The repository includes four primary plugins:

• aws-core – Located in plugins/aws-core/, this plugin covers service selection, CDK/CloudFormation, serverless computing, containers, storage, observability, billing, and SDK usage.
• aws-agents – Found in plugins/aws-agents/, this plugin provides skills for building AI agents using Amazon Bedrock and AgentCore.
• aws-data-analytics – Stored in plugins/aws-data-analytics/, this plugin handles data-lake architecture, ETL workflows, Athena queries, and Glue operations.
• aws-agents-for-devsecops – Located in plugins/aws-agents-for-devsecops/, this plugin supports incident investigation, code review, vulnerability scanning, and penetration testing using DevOps and Security Agents.

Skills and Rules Framework

Skills are self-contained markdown packages stored in the skills/ directory that provide step-by-step guidance, reference snippets, and executable scripts. Rules files (such as rules/aws-agent-rules.md) define policy-level guardrails including IAM condition keys, CloudWatch metrics, and CloudTrail audit logging requirements.

How AI Agents Interact With AWS

The workflow follows a discover-load-execute pattern. First, the agent identifies the appropriate plugin for the task at hand. Next, it loads the relevant skill files, which contain the specific instructions and code samples needed. Finally, the agent issues MCP calls or runs the provided Python scripts through the sandboxed environment to perform the AWS operation.

For example, when deploying a Lambda function, an agent might reference the aws-serverless skill from the aws-core plugin, then execute the provided deployment script that uses boto3 to interact with the Lambda API.

Security Guardrails and Compliance

The AWS Agent Toolkit for AWS implements multiple defense-in-depth mechanisms to ensure safe agent operations:

• IAM Condition Keys – Policies distinguish between agent-initiated and human-initiated actions, allowing fine-grained access controls.
• CloudWatch Metrics – Every request generates observability data for monitoring and anomaly detection.
• CloudTrail Audit Logging – Comprehensive audit trails capture all agent activities for compliance review.
• Pre-Installed Evaluations – Skills undergo verification checks before execution to validate reliability and safety.

These guardrails enable enterprises to adopt AI coding agents while maintaining strict security and compliance standards.

Implementation Examples

Installing the Core Plugin

For Claude Code users, add the foundational AWS capabilities:

/plugin install aws-core@claude-plugins-official

Installing the DevSecOps Plugin

To add security and operations capabilities:

/plugin marketplace add aws/agent-toolkit-for-aws
/plugin install aws-agents-for-devsecops
/reload-plugins
/aws-agents-for-devsecops:setup

Creating a Lambda Function

Using the aws-serverless skill from the aws-core plugin:

npx skills add aws/agent-toolkit-for-aws/skills
aws-serverless create-function \
  --function-name MyFunction \
  --runtime python3.11 \
  --handler app.handler \
  --zip-file fileb://my-function.zip

Running Sandboxed Python Scripts

The MCP server executes Python code in a restricted environment. For example, a skill might include a script like this to retrieve function configuration:

import boto3
import json

client = boto3.client('lambda')
response = client.get_function_configuration(FunctionName='MyFunction')
print(json.dumps(response, indent=2))

This script runs within the MCP server's sandbox, preventing unauthorized access to the underlying system while allowing legitimate AWS API calls.

Summary

• The AWS Agent Toolkit for AWS provides a plugin-based framework for AI agents to interact with AWS services through the Model Context Protocol.
• Four main plugins cover core AWS services (aws-core), AI agent development (aws-agents), data analytics (aws-data-analytics), and DevSecOps (aws-agents-for-devsecops).
• Skills are self-contained markdown packages with instructions and scripts, while Rules enforce security policies through IAM conditions and comprehensive logging.
• The MCP Server acts as a secure gateway, providing sandboxed Python execution and full AWS API coverage while maintaining audit trails via CloudTrail and CloudWatch.
• Enterprise teams can install specific plugins based on their use cases, allowing agents to deploy infrastructure, investigate security incidents, or analyze data while maintaining compliance.

Frequently Asked Questions

What is the AWS Agent Toolkit for AWS used for?

The AWS Agent Toolkit for AWS enables AI coding agents such as Claude Code, Codex, Cursor, and Kiro to safely manage AWS resources. It provides pre-built skills for common tasks like deploying Lambda functions, configuring S3 buckets, and investigating security incidents, all while enforcing enterprise security policies through integrated guardrails.

How does the MCP Server work in the AWS Agent Toolkit?

The MCP Server implements the Model Context Protocol to mediate between AI agents and AWS APIs. It authenticates agent requests, executes sandboxed Python scripts,provides real-time AWS documentation, and logs all activities to CloudTrail. This centralized approach ensures consistent security controls regardless of which specific plugin or skill the agent invokes.

What are the main plugins available in the AWS Agent Toolkit?

The repository includes four primary plugins: aws-core for general infrastructure and serverless computing, aws-agents for building Bedrock-based AI applications, aws-data-analytics for data lakes and ETL workflows, and aws-agents-for-devsecops for security operations and vulnerability management. Each plugin is stored in the plugins/ directory with its own skills and configuration.

How do I install plugins from the AWS Agent Toolkit?

Installation varies by agent platform. For Claude Code, use the /plugin install command followed by the plugin name and registry (e.g., /plugin install aws-core@claude-plugins-official). For other agents, you may need to add the marketplace repository first with /plugin marketplace add aws/agent-toolkit-for-aws, then install the specific plugin and reload the agent's plugin system.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}