Claude Code GitHub Action

Disambiguation (buyer-critical): this is claude-code-action, a GitHub Action you add to a repository's workflows — it is not the claude-code terminal CLI agent you run locally. They share the underlying Claude Code engine but are different products: the CLI is an interactive local coding agent, whereas this Action runs inside GitHub Actions CI, triggered by events (an @claude mention, issue assignment, or an automation prompt) and posts results back as PR comments/reviews. Vendor-described capabilities include interactive code and architecture assistance, automated PR analysis with improvement suggestions, implementing fixes/refactors/features, native GitHub comment integration, and progress tracking. It supports Anthropic API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry backends. The Action itself is free/MIT, but it runs on your own provider credentials, so real cost is provider token usage rather than a separate Action platform fee. Because it executes in CI with repository access, workflow permissions, untrusted prompt content, output visibility, and secret scoping are the primary risks to review. Telemetry behavior is provider-dependent: Claude Code's direct Anthropic API path enables operational metrics and Sentry error reporting by default, while Bedrock, Vertex AI, and Foundry defaults are generally off unless host-managed settings enable metrics; documented environment controls can disable telemetry, error reporting, and other non-essential traffic.