AWS recently released Agent Plugins for AWS, an open-source repository providing AI coding agents with specialized skills for architecting, deploying, and operating applications on AWS. The initial deploy-on-aws plugin transforms deployment workflows by accepting natural language commands like “deploy to AWS” and generating complete deployment pipelines with architecture recommendations, cost estimates, and infrastructure-as-code.
The plugin currently supports Claude Code and Cursor, following the recent marketplace launch of Cursor. The company positions agent plugins as a best practice for steering coding agents toward reliable outcomes without bloating model context. Rather than repeatedly pasting AWS guidance into prompts, developers can encode that guidance as reusable, versioned capabilities that agents invoke when relevant.
Marcelo S., AI product leader at AWS Agentic AI, framed the release in a LinkedIn post as addressing a fundamental scaling challenge:
For those who have spent years scaling tech organizations, the friction between writing code and deploying it at scale is often where velocity dies.
Moreover, he emphasized that leveraging the Model Context Protocol to bridge this gap ensures :
Our AI tools are not just generating text, but are actively operating within our technical ecosystems with precision and best-practice alignment.
AWS community builder Kento Ikeda, in a DEV.to blog post, suggests Agent Plugins represent more than automation. He argues that the significance lies in the layer at which the capability is integrated. Unlike CLI automation that optimizes command execution, the Agent Plugins systematizes the design process itself by explicitly defining the AWS workflow: analyze, recommend, estimate, generate, and deploy as structured agent capabilities. He writes:
CLI automation improves individual efficiency. Agent Plugins standardizes the design workflow.
This makes it valuable for organizational knowledge formalization and reproducible design patterns.
Agent plugins package multiple types of expertise artifacts together. A single plugin can include agent skills (structured workflows and best-practice playbooks), MCP servers (connections to external services and data sources), hooks (automation and guardrails on developer actions), and references (documentation and configuration defaults). The deploy-on-aws plugin uses three AWS MCP servers: AWS Knowledge for documentation and best practices, AWS Pricing for real-time cost estimates, and AWS IaC for CDK and CloudFormation guidance.
The plugin executes a five-step workflow:
- First, it analyzes the codebase to identify framework, database, and dependencies
- Second, it recommends optimal AWS services with rationale
- Third, it shows projected monthly costs using real-time pricing data
- Fourth, it generates CDK or CloudFormation infrastructure code
- Fifth, it deploys after user confirmation
AWS describes the workflow in a Developer Tool blog post with an Express.js REST API connected to PostgreSQL and serving a React frontend. After entering ‘deploy this Express app to AWS,’ the agent scanned the codebase, recommended AWS App Runner for the backend, Amazon RDS PostgreSQL for the database, and CloudFront plus S3 for the frontend, then provided cost estimates and generated complete infrastructure code, including Dockerfile and CI/CD workflows, before deploying everything. AWS claims this process took under 10 minutes versus hours of manual configuration.
Installation requires Claude Code or Cursor plus AWS CLI configured with appropriate credentials. In Claude Code, developers add the marketplace with /plugin marketplace add awslabs/agent-plugins, then install with /plugin install deploy-on-aws@awslabs-agent-plugins. Cursor users can install directly from the Cursor Marketplace or manually through settings.
AWS emphasizes that plugins are accelerators, not replacements for developer judgment. The launch post includes explicit guidance: always review generated code before deployment against security, cost, and resilience constraints; follow the principle of least privilege when configuring AWS credentials; and run security scanning tools on generated infrastructure code.
AWS plans to expand the plugin repository with additional capabilities for other AWS workflows in the coming weeks.
