Official toolkit lets AI coding agents build and deploy apps on AWS
If you have used AI coding tools like Claude Code or Cursor but had to manually search for AWS configurations, this repository automates that process. It consolidates AWS-supported MCP servers, skills, and plugins to help agents handle AWS services safely. You can now simply instruct the agent to call AWS APIs, search documentation, and execute scripts.
A unified toolkit following AWS Labs releases
MCP servers, skills, and plugins that were released separately by AWS Labs in 2025 have been integrated into this repository. AWS recommends using this unified version, which supports IAM condition keys to distinguish between agent and human actions, allowing for agent-specific policies. It also provides CloudWatch metrics and CloudTrail audit logs for all requests, enabling tracking and auditing of agent activity. Skills are evaluated end-to-end to ensure workflow completion.
Five specialized domains for agents
Plugins bundle the knowledge and configuration required for agents to perform specific AWS tasks.
- aws-agents Guides the construction of AI agents on AWS using Amazon Bedrock and AgentCore.
- aws-data-analytics Supports data lake, analytics, and ETL workflows using S3 Tables, AWS Glue, and Athena.
- aws-agents-for-devsecops Performs incident investigation, code review, UAT execution, vulnerability scanning, and penetration testing using AWS DevOps Agent and Security Agent.
- aws-startup-advisor Recommends minimal architectures suitable for startup stages and assists with migration from Azure, GCP, or Heroku to AWS, model invocation migration to Amazon Bedrock, and checking eligibility for AWS Activate credits.
Real-time AWS access via MCP servers
The AWS MCP server acts as the gateway for agents to communicate with AWS. It provides access to over 300 AWS services through a single authenticated endpoint. Agents can also execute Python scripts in isolated environments for complex multi-step tasks. Without authentication, agents can search for the latest AWS documentation, API references, and service features in real time, allowing them to make decisions based on current information. Enterprise control features include CloudWatch metrics, IAM context keys for per-agent policies, and CloudTrail audit logs.
Installation methods and requirements
Installation is done through the plugin marketplace for Claude Code, Codex, Cursor, and Devin. For example, in Claude Code, you can install it using the /plugin install aws-core@claude-plugins-official command. Other agents like Kiro or fx require separate setup for AWS MCP server configuration and skill installation. The MCP server configuration requires uv to be installed, and an account with local AWS credentials configured is needed for API calls and script execution. Credentials are not required if you only use documentation search or skill discovery.
Points to verify before use
This repository is published under the Apache-2.0 license. Since it is officially supported by AWS, you can trust its stability and maintenance. However, users must directly manage the costs and permission scopes associated with agent access to AWS resources. It is recommended to configure policies finely, such as granting read-only permissions to agents using IAM condition keys. Additionally, since the MCP server and skills operate independently, you can choose to use only one of them depending on your specific agent environment.
