How to Set Up the AWS Agent Toolkit Environment: Complete Setup Guide

To set up the AWS Agent Toolkit environment, install a supported agent plugin (Claude Code, Codex, Cursor, or Kiro) and configure the AWS MCP Server with your credentials to enable secure access to 300+ AWS APIs.

The aws/agent-toolkit-for-aws repository provides plugins, skills, and an AWS MCP Server that enable AI coding agents to interact with AWS services safely. Setting up the environment requires selecting your agent, installing the appropriate plugin from the repository, and configuring authentication through the MCP Server endpoint.

Choose Your AI Agent

The toolkit supports multiple AI coding agents through dedicated plugins. According to the [README.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/README.md) in the repository, you can use Claude Code, Codex, Cursor, Kiro, or any agent compatible with the AWS MCP Server.

• Claude Code – Uses slash commands for plugin management
• Codex – Integrates via the plugin marketplace
• Cursor – Supports Team Marketplace installations
• Kiro – Uses MCP configuration files and npm-based skill installation

Install the Plugin for Your Agent

Each agent has a specific installation path to load the AWS plugins and bundled skills.

Claude Code

For Claude Code, install the aws-core starter plugin using the slash command:

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

To update the marketplace index before installation:

/plugin marketplace update claude-plugins-official

Replace aws-core with aws-agents, aws-data-analytics, or aws-agents-for-devsecops to install specialized skill sets.

Codex

For Codex, add the toolkit to your marketplace and install interactively:

codex plugin marketplace add aws/agent-toolkit-for-aws

Then run /plugins in the Codex interface and select your desired plugin from the list.

Cursor

In Cursor, navigate to Settings → Plugins → Team Marketplaces and add the aws/agent-toolkit-for-aws repository as a Team Marketplace. Install your chosen plugin directly from the Plugins panel.

Kiro

For Kiro, first configure the MCP server in ~/.kiro/settings/mcp.json, then install skills via npm:

npx skills add aws/agent-toolkit-for-aws/skills

Other Agents

For agents not explicitly listed, configure the AWS MCP Server manually following the configuration below, then run the standard skill installation command:

npx skills add aws/agent-toolkit-for-aws/skills

Reference the repository's Getting Started guide in [README.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/README.md) for manual MCP configuration details.

Configure the AWS MCP Server

The AWS MCP Server authenticates your agent with AWS credentials and exposes over 300 AWS APIs through a single endpoint. As documented in the repository's [README.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/README.md), create a minimal MCP configuration:

{
  "mcpServers": {
    "aws": {
      "command": "uvx",
      "args": [
        "mcp-proxy-for-aws@1.6.3",
        "https://aws-mcp.us-east-1.api.aws/mcp",
        "--metadata", "AWS_REGION=us-west-2"
      ]
    }
  }
}

For Kiro users, write this configuration to ~/.kiro/settings/mcp.json:

cat > ~/.kiro/settings/mcp.json <<'EOF'
{
  "mcpServers": {
    "aws": {
      "command": "uvx",
      "args": [
        "mcp-proxy-for-aws@1.6.3",
        "https://aws-mcp.us-east-1.api.aws/mcp",
        "--metadata", "AWS_REGION=us-west-2"
      ]
    }
  }
}
EOF

The server uses uvx to run the MCP proxy, connecting to https://aws-mcp.us-east-1.api.aws/mcp with your specified region metadata.

Available Plugins and Skills

The aws/agent-toolkit-for-aws repository organizes capabilities into four primary plugins:

• aws-core – Starter plugin covering service selection, CDK/CloudFormation, serverless architectures, containers, observability, billing, SDK usage, and deployment workflows. See [plugins/aws-core/README.md](https://github.com/aws/agent-toolkit-for-aws/tree/main/plugins/aws-core).
• aws-agents – Skills for building AI agents on AWS using Amazon Bedrock and AgentCore.
• aws-data-analytics – Data lake, AWS Glue, Athena, and vector search workflows.
• aws-agents-for-devsecops – Incident investigation, code review, UAT, vulnerability scanning, and penetration testing capabilities. Detailed in [plugins/aws-agents-for-devsecops/README.md](https://github.com/aws/agent-toolkit-for-aws/tree/main/plugins/aws-agents-for-devsecops).

Individual skill packages are cataloged in [skills/README.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/skills/README.md).

Verification and User Guide

After installation, verify your setup by checking that the agent recognizes AWS-specific commands and can access the MCP Server endpoint. The repository maintains recommended policy files in [rules/aws-agent-rules.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/rules/aws-agent-rules.md) to guide agents on safe AWS usage.

For detailed walkthroughs on credential setup, MCP Server authentication, and advanced skill usage, consult the AWS Agent Toolkit User Guide.

Summary

• Install agent-specific plugins using commands like /plugin install for Claude Code or codex plugin marketplace add for Codex.
• Configure the AWS MCP Server using uvx and the endpoint https://aws-mcp.us-east-1.api.aws/mcp with your preferred AWS region.
• Select from four plugin tiers: aws-core (starter), aws-agents (Bedrock/AgentCore), aws-data-analytics, and aws-agents-for-devsecops.
• Reference key files: [README.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/README.md) for quick-start, [plugins/aws-core/README.md](https://github.com/aws/agent-toolkit-for-aws/tree/main/plugins/aws-core) for core capabilities, and [rules/aws-agent-rules.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/rules/aws-agent-rules.md) for safety policies.

Frequently Asked Questions

Which AI agents support the AWS Agent Toolkit?

The toolkit officially supports Claude Code, Codex, Cursor, and Kiro through dedicated plugins and configuration methods. Any AI coding agent that supports MCP (Model Context Protocol) servers can also integrate by manually configuring the AWS MCP Server endpoint and authentication.

What is the AWS MCP Server and why is it required?

The AWS MCP Server is a Model Context Protocol server that authenticates your agent with AWS credentials and exposes over 300 AWS APIs through a single secure endpoint. It acts as a proxy between your AI agent and AWS services, ensuring safe, controlled access without exposing raw credentials to the agent environment.

How do I change the AWS region for the MCP Server?

Modify the --metadata argument in your MCP configuration JSON. Change "AWS_REGION=us-west-2" to your desired region (for example, us-east-1 or eu-west-1) in the args array of the aws MCP server configuration. Restart your agent to apply the new region setting.

Where can I find documentation for specific AWS skills?

Detailed documentation for individual skill packages is available in [skills/README.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/skills/README.md). Plugin-specific guides are located in their respective directories under plugins/, such as [plugins/aws-core/README.md](https://github.com/aws/agent-toolkit-for-aws/tree/main/plugins/aws-core) for the core plugin.

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