Announced by Anthropic's Boris Cherny on X on September 14, 2026 ("Claude Mods are landing now"), with GitHub issue anthropics/claude-code#91870 serving as the coordination and demo thread rather than a formal documentation release. Unlike Agent Skills, MCP servers, hooks, and subagents — which all extend what the model can do, know, or call — early mods extend what the person using Claude Code sees and interacts with during a session, such as a playable Tetris game rendered inside the CLI, a breathing-exercise focus tool ("Mindful Claude" by @halluton), or a CI status panel built with RWX. As of publication there is no formal SDK, manifest format, or marketplace for mods.