Anthropic adds mods to Claude Code: TypeScript hooks inside the agent loop

Claude Code now supports mods, TypeScript hooks that run inside the agent loop and can rewrite prompts, gate or alter tool calls, decide permissions and redact secrets before they reach the model. For teams standardizing on Claude Code this is the control plane that was missing: policy as code, shipped with the repo, instead of a wiki page nobody reads. Expect a small ecosystem of shared mods within weeks.

Companies mentioned

AnthropicClaudeClaude Code