AWS

Use this Caplet when an agent needs live AWS account, Region, service, resource, IAM, operational, or AWS documentation context through the managed AWS MCP Server.

First Workflow

1. Start by confirming the intended account, Region, profile, service, and resource identifiers before querying broad AWS state.
2. Use documentation, skill, list, and describe operations to narrow ambiguous service behavior or resource matches before changing anything.
3. Inspect existing resource state, IAM context, dependencies, tags, and CloudTrail or service evidence before proposing operational changes.
4. For multi-account work, prefer named AWS profiles exposed through AWS_MCP_PROXY_PROFILES, and pass the intended profile on calls that support profile selection.
5. Use explicit Region names in requests when the target Region matters, especially if the runtime was not started with AWS_REGION.
6. Summarize the target account, Region, resource, and expected production effect before mutating resources.

Operate Carefully

• AWS operations can affect production infrastructure, data, security boundaries, billing, and compliance posture. Prefer read-only inspection before writes.
• Use least-privilege IAM roles, permission boundaries, and AWS MCP Server IAM condition keys where available.
• Confirm destructive or high-impact targets before deleting, replacing, scaling, deploying, rotating credentials, changing IAM, modifying network policy, or changing data stores.
• If credentials are missing or expired, refresh AWS CLI credentials and rerun the setup verification before retrying.
• Avoid this Caplet when the task only needs local IaC or application files; use the project workspace and deployment tooling for local configuration state.