Purpose

Agent Toolkit for AWS is designed to help AI coding agents build, deploy, and manage applications on AWS. It provides agents with the tools, knowledge, and guardrails needed to work with AWS services, and supports coding agents commonly used by developers, including Claude Code, Codex, Cursor, Kiro, and fx. This toolkit is the successor to the MCP servers, skills, and plugins in AWS Labs.

Key Capabilities

The toolkit includes plugins, Agent Skills, recommended configuration rules, and the AWS MCP Server.

Plugins

Plugins package AWS MCP Server configuration and agent skills into a single installation, and are currently available for Claude Code, Codex, and Cursor. Available plugins include:

aws-core: Covers service selection, CDK/CloudFormation, serverless, containers, storage, observability, billing, SDK usage, and deployment.
aws-agents: Covers building AI agents on AWS using Amazon Bedrock and AgentCore.
aws-data-analytics: Covers data lakes, analytics, and ETL workflows using S3 Tables, AWS Glue, and Athena.
aws-agents-for-devsecops: Used to investigate incidents, review code and perform UAT to confirm release readiness, scan code for vulnerabilities, and run penetration tests using AWS DevOps Agent and AWS Security Agent.
aws-startup-advisor: Covers architecture, cost, security, and AWS Activate guidance for startups, as well as migration to AWS from Azure, GCP, Heroku, and AI stacks such as OpenAI and Gemini.

Agent Skills

Agent Skills are curated packages of instructions and reference materials that help agents complete specific AWS tasks. Skills are loaded on demand, and agents discover and retrieve only the content relevant to the current task. The skills directory is skills/ in the repository.

Recommended Configuration Rules

Recommended project-level configuration files that guide agents on how to most effectivelyuse AWS, for example through the AWS MCP Server, discovering available skills, or searching documentation before taking action. See the rules/ directory for details.

AWS MCP Server

The AWS MCP Server is a managed server that gives agents access to AWS through the Model Context Protocol, providing:

Full AWS API coverage: Interact with 300+ AWS services through a single authenticated endpoint.
Sandboxed script execution: Agents can run Python scripts in an isolated environment for complex multi-step operations.
Real-time documentation access: Search and retrieve current AWS documentation, API references, and service capabilities without authentication.
Enterprise controls: Amazon CloudWatch metrics, IAM context keys for agent-specific policies, and AWS CloudTrail audit logs.

How to Use

Use directly in the terminal with the AWS CLI

Refer to the AWS CLI integration guide for setup, configuration, and usage.

Install plugins through the official Anthropic marketplace (Claude Code)

Plugins are available in the official Anthropic marketplace (claude-plugins-official), which is added to Claude Code installations by default. Use the command to install supported plugins:

aws-core:/plugin install aws-core@claude-plugins-official
aws-agents：/plugin install aws-agents@claude-plugins-official
aws-data-analytics：/plugin install aws-data-analytics@claude-plugins-official
aws-agents-for-devsecops：/plugin install aws-agents-for-devsecops@claude-plugins-official
aws-startup-advisor must be installed from this repository's marketplace:/plugin marketplace add aws/agent-toolkit-for-aws, then /plugin install aws-startup-advisor@agent-toolkit-for-aws Tip: If "Plugin not found" appears, first update the local marketplace index:/plugin marketplace update claude-plugins-official.

Codex

Run in the terminal:codex marketplace add aws/agent-toolkit-for-aws, then launch Codex and run /plugins to browse and install the aws-core plugin.

Cursor

In Settings → Plugins → Team Marketplaces → Add Marketplace → Import from Repo, add this repository as a team marketplace, pointing to aws/agent-toolkit-for-aws. When importing, Cursor indexes the plugins listed in .cursor-plugin/marketplace.json. Then open the Plugins panel to install the aws-core plugin (start here), or install aws-agents and aws-data-analytics as needed. Each plugin bundles AWS MCP Server configuration and agent skills.

Kiro

Setup consists of two separate parts: AWS MCP Server (for runtime AWS API access and documentation search) and local skills (for task-specific agent guidance). Skills do not require an MCP server, and the MCP server does not provide locally installed skills.

Add AWS MCP Server to the Kiro MCP configuration (.kiro/settings/mcp.json):

{
  "mcpServers": {
    "aws-mcp": {
      "command": "uvx",
      "args": ["mcp-proxy-for-aws@latest"]
    }
  }
}

Install skills from this repository:npx @aws/agent-toolkit@latest install-skills --agent kiro Skill files are installed to ~/.kiro/skills/ (global) or .kiro/skills/ (project-level). Each skill is a directory containing a SKILL.md file, optionally containing a references/ subdirectory. Kiro automatically discovers installed skills and activates them on demand when a task matches. Prerequisites: uv must be installed. Making API calls and executing scripts requires an AWS account with local credentials configured, but documentation search or skill discovery does not. For detailed setup, see the User Guide.

fx

Setup consists of the same two separate parts as Kiro.

Add AWS MCP Server from the terminal:fx mcp add aws-mcp -- uvx mcp-proxy-for-aws@latest. fx stores MCP servers in ~/.fx/mcp.json. You can also add it manually to ~/.fx/mcp.json; fx uses the mcp key instead of mcpServers, uses type to select the transport method, and a single command array holds the executable and its arguments. If a session is already open when you edit the file, run /mcp reload to pick up the changes.
Install skills:npx @aws/agent-toolkit@latest install-skills --agent fx. This installs the skill files to .fx/skills/ (project level); adding -g installs to ~/.fx/skills/ (global). fx also discovers skills already installed for other agents in .agents/skills/, .claude/skills/, and .codex/skills/. fx reads skill metadata at startup and loads skill instructions when you or an agent invokes them; type $ in the editor or run /skills to browse what is installed.

Other Agents

For other agents, configure the AWS MCP Server directly and install skills from this repository.

When to Use

Suitable for developers who need AI coding agents to build, deploy, and manage applications on AWS or perform tasks such as data analysis, DevSecOps, and startup architecture consulting. Supports common coding agents such as Claude Code, Codex, Cursor, Kiro, and fx.

Relationship with AWS Labs

In 2025, AWS began releasing MCP servers, skills, and plugins as part of AWS Labs. Agent Toolkit for AWS is the successor to these tools. Agent Toolkit for AWS is recommended because it provides the following key features:

IAM condition keys that distinguish agent actions from human actions, allowing you to write policies that apply only to agents. For example, even if the underlying IAM role can perform write operations, you can write a policy that allows only read-only operations through the MCP server.
Per-request CloudWatch metrics and CloudTrail audit logs, allowing you to monitor and audit coding agent activity.
Agent skills that have undergone comprehensive end-to-end evaluation, ensuring successful completion of workflows. AWS Labs MCP servers, skills, and plugins will continue to work and accept contributions, and over time the best tools in AWS Labs will move to Agent Toolkit for AWS.

Limitations

The materials do not provide detailed information about fees, pricing, or specific platform limitations. Usage requires installing uv (for the local skills portion of Kiro and fx), and making API calls and executing scripts requires an AWS account with local credentials configured. For detailed operations, available tools, authentication, and supported regions, see the AWS MCP Server documentation.

License

This project uses the Apache-2.0 license.

agent-toolkit-for-aws

About this tool