AWSAgent Toolkit for AWS gives AI coding agents the tools, knowledge, and guardrails to build, deploy, and manage applications on AWS. It has three parts: the managed AWS MCP Server, a set of agent skills, and recommended rules files, with plugins that bundle the server configuration and skills into a single install.
Plugins are available for Claude Code, Codex, and Cursor. Other coding agents, including Kiro and fx, configure the AWS MCP Server directly and install the skills separately. The two parts work independently: skills do not require the MCP server, and the server does not serve locally installed skills. The toolkit is also reachable from the AWS CLI.
The toolkit is the successor to the MCP servers, skills, and plugins released under AWS Labs in 2025. Compared with those, it adds IAM condition keys that tell agent actions apart from human actions (so a policy can, for example, allow only read-only actions through the MCP server), CloudWatch metrics and CloudTrail audit logs for every request, and skills that have gone through end-to-end evaluations.
| Plugin | Covers |
|---|---|
| aws-core | Service selection, CDK and CloudFormation, serverless, containers, storage, observability, billing, SDK usage, and deployment |
| aws-agents | Building 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 | Incident investigation, code review, release-readiness UAT, vulnerability scans, and penetration tests with AWS DevOps Agent and AWS Security Agent |
| aws-startup-advisor | Startup architecture, cost, security, AWS Activate guidance, and migrations from Azure, GCP, Heroku, OpenAI, and Gemini stacks |
Features
- AWS MCP Server: one authenticated endpoint reaches the APIs of more than 300 AWS services
- Sandboxed scripts: agents run Python scripts in an isolated environment for multi-step operations
- Documentation search: current AWS documentation, API references, and service capabilities, searchable without authentication
- Agent skills: curated instructions and reference material that agents load on demand when a task matches
- Rules files: project-level configuration that tells agents to use the MCP server, discover skills, and search documentation before acting
- Coding-agent plugins: single-install bundles for Claude Code, Codex, and Cursor
- Agent-aware IAM: condition keys let policies apply only to agent actions
- Auditing: CloudWatch metrics and CloudTrail logs record every agent request
