GAOfficial, AWS-supported. The successor to the AWS Labs tools.

Help AI coding agents build on AWS

Agent Toolkit for AWS gives Claude Code, Codex, Cursor, Kiro, and any MCP-compatible agent the tools, curated skills, and guardrails to ship production-ready infrastructure — instead of trial-and-error against outdated training data.

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

Installs the aws-core plugin from the official marketplace.

Apache-2.0 licensedNo additional chargeWorks with the agents you already use

claude-code — aws-core plugin active

you > Create an S3 bucket with versioning enabled and a lifecycle policy that transitions objects to Glacier after 90 days.

skill loading s3-storage-setup tested procedure · loaded on demand

mcp aws · call_aws s3api create-bucket · us-west-2

mcp aws · call_aws put-bucket-versioning · put-lifecycle

iam guardrail · aws:CalledViaAWSMCP

Agent-initiated calls scoped by IAM context key. Every request logged to CloudTrail with CloudWatch metrics.

agent > Done. Bucket app-assets-prod created with versioning on and a 90-day Glacier transition. Followed the AWS Well-Architected storage guidance.

MCP 300+ servicesskills loaded on demandCloudTrail full audit

300+

AWS services reachable through one endpoint

15,000+

API actions available via the MCP Server

2.4k+

GitHub stars · Apache-2.0, open source

No additional charge — pay only for AWS you use

What's included

Tools, knowledge, and guardrails — in three parts.

AWS MCP Server

A managed, remote server that gives agents secure access to AWS through the Model Context Protocol — one authenticated endpoint.

Full coverage — 300+ services, 15,000+ API actions, no local CLI
Sandboxed Python execution, isolated from your filesystem and network
Real-time AWS documentation and API-reference search
Enterprise controls — CloudWatch metrics, IAM context keys, CloudTrail

Agent Skills

Curated packages of instructions, scripts, and reference material that guide agents through complex AWS tasks. Loaded on demand to keep context small.

Service decision guides with opinionated tradeoff analysis
Tested step-by-step procedures for common workflows
Troubleshooting guides with root-cause analysis
Discoverable at runtime or pre-installed — all open source

Agent Plugins

Easy-install bundles that combine the MCP Server configuration and a curated set of skills for your agent — with automatic updates.

One install for Claude Code, Codex, or Cursor
Skills and MCP config update automatically
Rules files tell agents how to use AWS most effectively
Open source in the agent-toolkit-for-aws repo

Official plugins

Pick the bundle that matches the job.

aws-coreStart here

Service selection, CDK/CloudFormation, serverless, containers, storage, observability, billing, SDK usage, and deployment.

aws-agents

Build AI agents on AWS with Amazon Bedrock and AgentCore.

aws-data-analytics

Data lake, analytics, and ETL workflows with S3 Tables, AWS Glue, and Athena.

aws-agents-for-devsecops

Investigate incidents, review code, run UAT and pen tests with the AWS DevOps and Security agents.

Plugins are available for Claude Code, Codex, and Cursor. For other agents, configure the AWS MCP Server directly and install skills from the repo.

Get started

Set it up in the agent you already use.

Prerequisites: install uv for MCP setups. An AWS account with local credentials is needed for API calls — not for docs search or skill discovery.

Claude Code

Plugins live on the official Anthropic marketplace, added to Claude Code by default. Install aws-core to start.

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

Codex

Add the marketplace from your terminal, then launch Codex and run /plugins to install aws-core.

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

Kiro & other MCP agents

Add the AWS MCP Server to your MCP config, then install skills. Pin the proxy version for reproducible builds.

{
  "mcpServers": {
    "aws": {
      "command": "uvx",
      "args": [
        "mcp-proxy-for-aws@1.6.4",
        "https://aws-mcp.us-east-1.api.aws/mcp",
        "--metadata", "AWS_REGION=us-west-2"
      ]
    }
  }
}

# then install skills
npx skills add aws/agent-toolkit-for-aws/skills

Try it out

Point your agent at AWS and ask. Skills load on demand and every call is scoped by IAM and logged to CloudTrail.

Ask your agent:
"Create an S3 bucket with versioning enabled
 and a lifecycle policy that moves objects
 to Glacier after 90 days."

Read the user guide →View on GitHub

FAQ

Questions, answered.

MCP is an open standard for connecting AI agents to external tools and data sources. It gives agents a standard way to discover tools, call them with structured inputs, and receive structured outputs. It's maintained by the Agentic AI Foundation under the Linux Foundation, with AWS as a founding member.