What's New

AWS has launched the Agent Toolkit for AWS, a production-ready suite of tools designed to help AI coding agents build on AWS more accurately, efficiently, and securely. The toolkit includes over 40 curated agent skills, a generally available fully-managed MCP (Model Context Protocol) server, and three pre-packaged agent plugins (AWS Core, AWS Data Analytics, and AWS Agents). It serves as the official successor to the experimental MCP servers, plugins, and skills previously available on AWS Labs.

How It Works

• The Agent Toolkit operates across three integrated layers. Agent Skills are validated, up-to-date procedural guides that instruct coding agents on how to correctly perform specific AWS tasks — such as authoring CloudFormation templates, configuring data pipelines, or building serverless applications — replacing ad-hoc inference from general training data with curated best-practice workflows.
• The AWS MCP Server (now GA) is a fully-managed server implementing the Model Context Protocol that brokers interactions between coding agents and any AWS service; it enforces IAM-based guardrails to restrict agent actions, provides observability via Amazon CloudWatch and AWS CloudTrail, executes multi-step operations in a sandboxed environment, and dynamically retrieves up-to-date AWS documentation to keep agents current. Agent Plugins bundle the MCP server with domain-specific skill sets into a single installable package: AWS Core targets full-stack application development, AWS Data Analytics targets pipeline and BI engineering workflows, and AWS Agents targets AI engineers building on Amazon Bedrock AgentCore.

Why It's Important

• AI coding agents working with AWS have historically suffered from three compounding problems: hallucinating incorrect API calls or configurations due to stale training data, accumulating high token costs from trial-and-error on complex multi-service workflows, and lacking enterprise-grade governance controls that make production deployment risky.
• The Agent Toolkit directly addresses all three: skills reduce errors by providing validated procedures, the MCP server's documentation retrieval keeps agents current without retraining, and IAM guardrails plus CloudTrail/CloudWatch observability give platform and security teams the control and audit trail needed to confidently deploy agents in regulated or production environments.
• For organizations investing in AI-assisted development, this significantly lowers the barrier to trustworthy, cost-effective agentic workflows on AWS.

How It's Different

• The Agent Toolkit for AWS supersedes the AWS Labs ecosystem of MCP servers, plugins, and skills, which were experimental, community-oriented, and lacked production-grade governance.
• The key differentiators are: (1) Rigorously evaluated skills — each skill is tested for accuracy and reliability rather than being community-contributed without formal validation; (2) Fully-managed MCP server — previously, users had to self-host or manage MCP server infrastructure, whereas the new server is AWS-managed with built-in IAM guardrails, sandboxed execution, and native observability integrations; (3) Opinionated plugin bundles — rather than assembling tooling piecemeal, developers get role-specific, single-install plugins; and (4) Enterprise governance — CloudTrail integration and IAM-scoped action controls are first-class features, not afterthoughts.
• Compared to generic MCP implementations or third-party agent frameworks, the toolkit is deeply integrated with AWS service APIs and documentation pipelines.

When to Prefer It

• Choose the Agent Toolkit for AWS when your team is actively using AI coding agents (e.g., Amazon Q Developer, GitHub Copilot, Cursor, or custom agents) to build, deploy, or manage AWS infrastructure and applications, and you need production-grade reliability and governance.
• It is especially valuable in enterprise environments where security teams require audit trails and scoped permissions for any automated agent actions.
• The AWS Data Analytics plugin is the right choice for data engineering teams automating pipeline creation and BI workflows, while the AWS Agents plugin is purpose-built for AI engineers building multi-agent systems on Bedrock AgentCore.
• If you are currently using AWS Labs MCP servers or skills, migrating to the Agent Toolkit is the recommended path forward for stability and support.
• For purely experimental or niche use cases not yet covered by the 40+ available skills, AWS Labs tooling remains a valid complement in the interim.

Availability

• The Agent Toolkit for AWS, including the AWS MCP Server, is generally available (GA) as of May 6, 2026.
• The toolkit is available at no additional charge — customers pay only for the underlying AWS resources consumed by their agents.
• More than 40 skills are available at launch across infrastructure-as-code, storage, analytics, serverless, containers, and AI services; additional skills covering databases, networking, and IAM are planned for release in the coming weeks.
• Three agent plugins (AWS Core, AWS Data Analytics, AWS Agents) are available now.
• Regional availability details are not explicitly stated in the announcement, but as a managed service integrated with standard AWS APIs, it is expected to be broadly accessible.
• The toolkit and its skills and plugins can be explored and installed via the AWS Agent Toolkit GitHub repository, with onboarding available through the official Quick Start guide.
• Existing AWS Labs MCP servers and skills remain available and will be progressively migrated into the Agent Toolkit over time.