Get AWS Native SQL Help With Kiro

1. If your team already has Amazon Q Developer's IDE plugin installed and working, you can keep using it until AWS's support window closes, but don't build new workflows around it since it's being phased out.
2. If you're setting this up fresh, or migrating, go with Kiro as AWS's current recommended path. Cursor, Claude Code, and GitHub Copilot are also valid AWS-aware alternatives if your team has already standardized on one of those.

1. Download and install Kiro from AWS's or Kiro's official site.
2. Sign in to associate it with your AWS account.

What you should see: An editor interface with an AI panel or chat available, similar in shape to other AI-enabled editors you may have used.

Troubleshooting: If your company restricts installing new applications, check with your platform or security team, since AWS tooling like this is often already vetted for engineering roles even if the specific product is new.

1. Configure the AWS credentials Kiro uses to be a role scoped to what you actually need, read access to the specific Redshift cluster or Athena workgroup you work in, not your full account's admin permissions.
2. If you don't already have a role like this, ask whoever manages IAM on your team to create one before you connect anything. This is a five-minute request that avoids a much bigger problem later if a credential ever leaks through a shared file or chat log.

Why this matters: An AI assistant with the same permissions as an account admin is a bigger blast radius than the convenience is worth. Scope it down before you connect it, not after something goes wrong.

1. Open a query you're debugging, or paste one directly.
2. Ask Kiro to explain the query plan and where the time is going.
3. Review the explanation against what you already know about the table's size and structure before trusting a specific recommendation.

What you should see: An explanation referencing your actual AWS resource configuration (cluster size, table statistics) rather than generic advice about query optimization in the abstract.

1. Describe what you need in plain language: "write a query against the orders table that returns daily totals for the last 30 days."
2. Review the generated SQL against the actual table structure before running it, especially column names and types, since generated SQL can reference a column that doesn't exist or has a different type than expected.

1. Never paste real customer records, account numbers, or other sensitive values into a prompt, even as an example of what a query should return. Describe the shape of the data instead ("a column with dollar amounts," not an actual dollar amount tied to a real customer).
2. Data pulled back from Redshift or Athena as sample rows for review may contain PII depending on the table. Treat anything the AI shows you from a live query result with the same care you'd give a query result in any other tool, and don't forward it outside your team without checking your data handling policy first.