# Agent Toolkit for AWS: Official MCP Servers and Skills for AI Coding Agents

> Agent Toolkit for AWS is the official, AWS-supported collection of MCP servers, agent skills, and plugins that give AI coding agents the tools and knowledge to build, deploy, and manage applications on AWS. It works with Claude Code, Codex, Cursor, Kiro, and fx, and covers services from CDK and CloudFormation to Amazon Bedrock and data analytics pipelines.

**aws/agent-toolkit-for-aws** — Official, AWS-supported MCP servers, skills, and plugins to help AI agents build on AWS

- Repository: https://github.com/aws/agent-toolkit-for-aws
- Stars: 2,762 · Forks: 337
- Language: Python
- License: Apache-2.0
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/aws-agent-toolkit-for-aws

## What the Agent Toolkit for AWS Does

The Agent Toolkit for AWS gives AI coding agents structured knowledge and runtime access to AWS services. It addresses a specific problem: AI agents working on AWS code often lack reliable, up-to-date context about service selection, CDK patterns, IAM constraints, and deployment pipelines. The toolkit provides that context as agent skills (static instruction files) and as MCP servers (runtime tools that give agents direct access to AWS APIs, documentation search, and sandboxed script execution).

The toolkit is maintained by AWS and is described in the README as officially supported. It targets developers who already use one of the supported coding agents: Claude Code, Codex, Cursor, Kiro, or fx. The README describes the toolkit as working with agents developers already use, not as a standalone tool.

The plugin collection is published under Apache-2.0.

## Available Plugins and What Each Covers

The toolkit provides five plugins.

The aws-core plugin is the starting point. It covers service selection guidance, CDK and CloudFormation patterns, serverless architecture, containers, storage, observability, billing, SDK usage patterns, and deployment workflows.

The aws-agents plugin covers building AI agents on AWS using Amazon Bedrock and AgentCore. It is separate from aws-core because the AI agent development workflow has distinct patterns and services.

The aws-data-analytics plugin covers data lake design, analytics pipelines, and ETL workflows with S3 Tables, AWS Glue, and Athena.

The aws-agents-for-devsecops plugin covers incident investigation, code review for release readiness, code vulnerability scanning, and penetration testing using AWS DevOps Agent and AWS Security Agent.

The aws-startup-advisor plugin covers startup-focused architecture guidance, cost optimisation, security, AWS Activate programme context, and migration guidance from Azure, GCP, Heroku, and AI stacks including OpenAI and Gemini.

## Installing into Claude Code

The aws-core and aws-agents plugins are available on the official Anthropic marketplace, which the README states is added to Claude Code installations by default. The install commands are:

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

```
/plugin install aws-agents@claude-plugins-official
```

If the marketplace index is stale and returns a plugin-not-found error, update it first:

```
/plugin marketplace update claude-plugins-official
```

The aws-data-analytics plugin also installs from the official marketplace:

```
/plugin install aws-data-analytics@claude-plugins-official
```

The aws-agents-for-devsecops plugin can install from either the official marketplace or from this repository's marketplace directly:

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

After installing aws-agents-for-devsecops, run /aws-agents-for-devsecops:setup to complete configuration. The aws-startup-advisor plugin is only available from this repository's marketplace, not the official Anthropic one.

## Installing into Codex, Cursor, Kiro, and fx

For Codex (version 0.122 or later), add the marketplace and browse plugins:

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

Then launch Codex and run /plugins to browse and install aws-core.

For Cursor, add the repository as a team marketplace from Settings, pointing at aws/agent-toolkit-for-aws. Cursor indexes the plugins listed in .cursor-plugin/marketplace.json. Install aws-core first from the Plugins panel.

For Kiro, the setup has two independent parts. The first is the AWS MCP Server, configured in .kiro/settings/mcp.json:

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

The second part is the skills, installed with:

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

For fx, the MCP server adds to ~/.fx/mcp.json and skills install with:

```bash
npx skills add aws/agent-toolkit-for-aws/skills -a fx
```

Both Kiro and fx require uv to be installed. An AWS account with credentials configured locally is required for API calls and script execution, but not for documentation search or skill discovery.

## The AWS MCP Server: Runtime API Access and Documentation Search

The MCP server component is separate from the skill files. While skills provide static instruction sets that the agent reads, the MCP server gives the agent runtime access to AWS APIs, sandboxed script execution, and real-time documentation search.

The MCP server proxies through the AWS CLI, which means it uses the locally configured AWS credentials and region. The sandboxed script execution capability lets the agent run AWS CLI commands and small scripts in a controlled environment as part of its task.

For Kiro and fx, the skills and the MCP server work independently. The README states explicitly that skills do not require the MCP server and the MCP server does not serve locally installed skills. The combination provides the most complete capability: the skills guide the agent's reasoning, and the MCP server gives it live access to verify current service states, search current documentation, and execute AWS operations.

For Claude Code users, each plugin bundles the AWS MCP Server configuration alongside the agent skills, so the separation is less visible.

## Repository Layout: Plugins, Skills, Rules, and Tools

The repository root contains four substantive directories. The plugins/ directory holds the plugin manifests and content for each named plugin. The skills/ directory holds skill files compatible with agents that use the npx skills add installation method. The rules/ directory holds rule files for agents that support rule-based guidance. The tools/ directory holds additional tooling scripts.

The .agents/ directory holds agent-agnostic configuration. The .claude-plugin/ directory holds Claude Code specific plugin metadata. The .cursor-plugin/ directory holds Cursor-specific plugin manifests, including .cursor-plugin/marketplace.json. The setup-instructions/ directory holds per-agent setup documentation.

The mise.toml file specifies the tool versions used in the repository. The repository follows Apache-2.0 licensing, with a NOTICE file for attribution.

## Limitations and What the Toolkit Does Not Provide

The toolkit provides guidance and tooling for AWS but does not substitute for AWS documentation or Solutions Architect review on complex architectures. The skills encode known patterns and best practices as of their last update, but AWS service behaviour, IAM policy syntax, and CDK constructs change with service releases. The README does not document how frequently skill content is updated relative to AWS service changes.

The MCP server requires an AWS account with configured local credentials for API calls. Developers working in environments without direct AWS CLI access, such as certain CI systems or corporate networks with strict outbound controls, will find the runtime API access unavailable. The documentation search capability works without AWS credentials, as the README notes.

The aws-startup-advisor plugin is not on the official Anthropic marketplace; it requires adding this repository's marketplace first. The aws-agents-for-devsecops plugin requires a /reload-plugins command after installation, which is an extra step compared to the other plugins.

A comparable alternative is the AWS Toolkit for VS Code or JetBrains, which provides IDE-native AWS resource browsing, CloudFormation linting, and SAM debugging. The VS Code toolkit operates within the IDE rather than through an AI agent conversation, so it suits developers who prefer direct GUI interaction over agent-mediated workflows.

## Conclusion

Agent Toolkit for AWS is the correct starting point for developers who use Claude Code, Codex, Cursor, Kiro, or fx and want structured AI guidance on AWS architecture, deployment, security, and data analytics. The aws-core plugin covers the broadest range of AWS topics and is the right first install. The aws-agents plugin adds specific Bedrock and AgentCore coverage for teams building AI agents on AWS. Before installing, confirm that your agent version supports the plugin or skill format the toolkit uses: Claude Code users need the official Anthropic marketplace available by default, Codex requires version 0.122+, and Kiro and fx both require uv. The last push was on 2026-09-26.

## FAQ

### What is the agent toolkit for AWS?

Agent Toolkit for AWS is the official, AWS-supported collection of MCP servers, agent skills, and plugins that give AI coding agents structured knowledge and runtime access to AWS services. It works with Claude Code, Codex, Cursor, Kiro, and fx.

### What does the AWS toolkit do?

The Agent Toolkit for AWS provides five plugins covering core AWS services (aws-core), AI agent development on Bedrock (aws-agents), data analytics pipelines (aws-data-analytics), DevSecOps workflows (aws-agents-for-devsecops), and startup architecture guidance (aws-startup-advisor). An MCP server component also gives agents runtime access to AWS APIs and documentation search.

### what is agent toolkit for aws

It is an officially supported AWS repository of MCP servers, skills, and plugins that integrate with AI coding agents to help developers build, deploy, and manage AWS applications. The aws-core plugin is the recommended starting point and installs into Claude Code with /plugin install aws-core@claude-plugins-official.

## Sources

- [aws/agent-toolkit-for-aws on GitHub](https://github.com/aws/agent-toolkit-for-aws)
- [Issues](https://github.com/aws/agent-toolkit-for-aws/issues)
- [License: Apache-2.0](https://github.com/aws/agent-toolkit-for-aws/blob/main/LICENSE)
- [README](https://github.com/aws/agent-toolkit-for-aws/blob/main/README.md)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/aws-agent-toolkit-for-aws