AWS CLI Agent Toolkit Adds Skills and MCP Setup for Coding Agents

AWS is packaging cloud expertise for coding agents

AWS published a new Agent Toolkit for AWS workflow on August 26, 2026 that lets developers configure supported AI coding agents from the AWS CLI. The goal is to give agents current AWS service guidance plus a standardized, auditable route to AWS APIs instead of relying only on whatever cloud knowledge is embedded in the model.

The toolkit requires AWS CLI version 2.35.0 or newer and supports popular coding agents including Kiro, Claude Code, Codex and Cursor, with additional supported clients discoverable from the CLI.

One command installs skills and configures MCP

The main setup command is aws configure agent-toolkit. AWS says the wizard detects supported coding agents on the local machine, lets the user choose which ones to configure, installs a default set of AWS skills and configures the AWS MCP Server connection.

For non-interactive environments such as CI or development containers, AWS documents aws configure agent-toolkit --yes to accept defaults without prompts.

AWS skills are maintained knowledge packages

The toolkit includes a catalog of Markdown-based AWS skills maintained by AWS service teams. These skills are intended to give coding agents current guidance about services and common development workflows. Developers can search, add, update and remove skills as a project changes.

That addresses a common weakness in AI coding workflows: a model may know the general shape of an AWS service but miss recently released APIs, renamed capabilities or current security practices. A service-team-maintained skill can update faster than the underlying foundation model.

The AWS MCP Server is the action layer

The second major component is the AWS MCP Server, which AWS describes as a secure and auditable way for agents to list resources, search documentation and call AWS APIs using existing IAM credentials. That means the toolkit separates knowledge from execution: skills provide curated context, while MCP provides controlled access to live cloud capabilities.

Existing IAM permissions still matter. Installing the toolkit does not grant an agent unlimited AWS access; actions remain constrained by the credentials and policies available to the configured client.

Why this matters for agentic development

Coding agents increasingly scaffold infrastructure, select services and modify cloud configurations. AWS is turning the CLI into a distribution point for both agent knowledge and tool connectivity, which can reduce one-off setup across different coding assistants.

The broader signal is important: cloud vendors are beginning to ship agent-specific operational context as a maintained product layer, rather than expecting each model or IDE vendor to independently encode fast-changing cloud knowledge.