# Claude Code Base Action

A GitHub Action that installs and runs Claude Code inside a workflow so a team can build custom automation on top of it; it enforces no trust boundary of its own, so untrusted-input workflows should use the separate claude-code-action instead.

This GitHub Action is a thin wrapper that installs and runs Claude Code with the inputs a workflow provides, so a team can build a custom automation on top of Claude Code inside GitHub Actions. Running it in a directory is equivalent to running Claude Code there: it reads project configuration such as .claude/, CLAUDE.md and .mcp.json from the working directory.

Because it does not enforce any trust boundary on its own, the caller is responsible for making sure the working directory and prompt are trusted. Workflows that process untrusted input such as issues, fork pull requests or external comments should use the separate anthropics/claude-code-action instead, which adds actor permission checks and restores project configuration from the base ref.

Authentication can use a static API key, an OAuth token, or Workload Identity Federation, which exchanges a workflow's GitHub OIDC token for a short-lived access token.

Website: https://github.com/anthropics/claude-code-base-action