Prerequisites

Before you install eksreview, make sure your machine and AWS access are set up as described below. If you already work with EKS, you probably have most of this in place.

Requirement	Notes
Operating system	macOS or Linux. On Windows, run it under WSL, since the install.sh and eksreview launchers are bash scripts.
Python 3.10+	3.11 or 3.12 recommended
uv	Required for the bundled MCP server. install.sh installs it for you if it's missing; you can also install it yourself with the install guide.
AWS access	Credentials with permissions for EKS, EC2, IAM, STS, and Bedrock
Amazon Bedrock model access	Claude Opus and/or Sonnet enabled in your Bedrock region. By default the agent uses global cross-region inference profiles, which work from commercial regions worldwide. To pin a region, set MODEL_ID to a regional inference profile.
Cluster access for your IAM identity	Your IAM principal mapped into each cluster you review, via an EKS access entry (recommended) or the aws-auth ConfigMap

Supported clusters: any Amazon EKS cluster, on any supported Kubernetes version, in standard mode or EKS Auto Mode. On Auto Mode clusters the Karpenter and Cluster Autoscaler checks that don't apply are automatically reported as "not applicable" rather than failing. The upgrade-readiness checker pulls its API-deprecation data live, so it stays current with new Kubernetes releases.