Is the AWS Agent Toolkit Open Source? Yes—Here’s the Proof

The AWS Agent Toolkit is fully open source under the Apache 2.0 license, with its complete source code available in the public GitHub repository aws/agent-toolkit-for-aws.

The AWS Agent Toolkit is an open-source project that provides plugins, skills, and safety rules for building AI agents that interact with AWS services. Because it is hosted publicly on GitHub and released under a permissive Apache 2.0 license, anyone can clone the repository, inspect the implementation, build it locally, or contribute improvements.

Open Source License and Repository Structure

The project declares its open-source status explicitly in the LICENSE file at the root of the repository. The file states: “This project is licensed under the Apache-2.0 License” (see LICENSE). This license grants users the right to use, modify, and distribute the code for commercial and non-commercial purposes, provided the license terms and copyright notices are preserved.

Key source files that demonstrate the open nature of the project include:

• README.md – Contains the project overview, quick-start guides, and architecture documentation
• LICENSE – The Apache 2.0 license text
• plugins/ – Bundled configurations for the MCP Server
• skills/ – Markdown-based skill definitions and reference documentation
• rules/aws-agent-rules.md – Safety and governance guardrails

Core Architecture and Public Source Files

The toolkit’s architecture is organized around three primary concepts, all of which are implemented as publicly readable files in the repository.

Plugins

Plugins bundle the Model Context Protocol (MCP) Server configuration and skill sets for specific agent types. The repository contains multiple plugin packages under the plugins/ directory:

• plugins/aws-core/ – Core AWS infrastructure skills
• plugins/aws-agents/ – Agent-specific configurations
• plugins/aws-data-analytics/ – Data analytics workflow bundles

Each plugin ships a pre-configured MCP Server endpoint and curated skills that map to AWS workflows such as serverless deployment, container management, and DevSecOps.

Skills

Skills are stored in the skills/ directory and are lazily loaded at runtime. Each skill package contains:

• A SKILL.md file defining metadata, description, and parameters
• Reference markdown files holding step-by-step logic

For example, the Lambda creation skill includes detailed instructions in skills/core-skills/aws-serverless/references/lambda.md (see [lambda.md](https://github.com/aws/agent-toolkit-for-aws/blob/main/skills/core-skills/aws-serverless/references/lambda.md)).

Rules

The rules/aws-agent-rules.md file contains project-level guardrails that enforce safe AWS usage. These rules specify that only the MCP Server may invoke AWS APIs, require CloudTrail audit logging for every request, and limit cost-impact actions.

How to Access and Use the Open Source Code

Because the AWS Agent Toolkit is open source, you can install plugins directly from the public repository or clone the source to customize the behavior.

Install the Core Plugin

You can install plugins in various agent environments using the public repository URI:

# Claude Code (Marketplace)

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

# Codex (terminal)

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

# then in the UI run /plugins → install aws-core

# Cursor (team marketplace import)

# Follow Settings → Plugins → Team Marketplaces → Add Marketplace → Import from Repo

# (point to https://github.com/aws/agent-toolkit-for-aws)

Install the DevSecOps Plugin Suite

# Add the repository to the marketplace

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

# Install the dev-sec-ops bundle

/plugin install aws-agents-for-devsecops
/reload-plugins

# Run the setup command

/aws-agents-for-devsecops:setup

Load and Invoke Skills Locally

You can load specific skills using the npx helper and invoke them directly:

# Use the npx helper to add all skills (or a subset)

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

# Then invoke a skill, e.g., create a Lambda function

/aws-core:aws-serverless:create-lambda \
  --function-name myFunction \
  --runtime nodejs18.x \
  --handler index.handler

Summary

• The AWS Agent Toolkit is definitively open source, released under the Apache 2.0 license.
• The full source code resides in the public GitHub repository aws/agent-toolkit-for-aws.
• All components—including plugins (plugins/), skills (skills/), and safety rules (rules/aws-agent-rules.md)—are publicly accessible and modifiable.
• The MCP Server acts as the runtime interface, while the toolkit itself consists of configuration and documentation files that can be customized for specific use cases.

Frequently Asked Questions

Is the AWS Agent Toolkit free to use?

Yes. As an open-source project under the Apache 2.0 license, the AWS Agent Toolkit is free to use, modify, and distribute. You can use it in commercial applications without licensing fees, provided you include the required license and copyright notices.

Where can I view the AWS Agent Toolkit license?

The license is available in the LICENSE file at the root of the repository (LICENSE). The README.md also contains a dedicated License section that confirms the Apache 2.0 status.

Can I contribute to the AWS Agent Toolkit?

Yes. Since the repository is public on GitHub, you can fork the project, submit pull requests, and open issues. The open-source nature encourages community contributions to plugins, skills, and documentation.

What components of the AWS Agent Toolkit are open source?

All components are open source, including:

• Plugin definitions in plugins/aws-core/, plugins/aws-agents/, and other directories
• Skill definitions in skills/ with their markdown reference files
• Safety rules in rules/aws-agent-rules.md
• Documentation and configuration files

The only external dependency is the MCP Server binary, which is published separately on PyPI, while the toolkit configurations that interface with it are fully public.

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