Claude Code just got mods

Claude Code just got mods. Now you can change how Claude Code itself works, with a few lines of TypeScript.

Anthropic launched them on 1 Oct 2026. Think of game mods: the game stays the same, your mod changes how it plays.

In simple words

• Every time Claude Code does something (runs a tool, draws the screen), it fires an event
• A mod is a small function that catches it and can change, block or replace it

Key features

• Chat to chat: send a message from one session to another. Finish in chat A, hand the result to chat B
• Smarter prompts: rewrite your prompt before the model sees it
• Safety guard: hold rm -rf or a force push until you click Proceed
• Secret filter: hide keys in tool output before Claude reads them
• Your own screen: panels and buttons, like a live "how full is my context" meter
• Instant commands: a /command runs your code at once, even while Claude works
• Model switch: send a request to a different model
• Swap built-ins: /diff is already a mod you can replace
• Self-build: ask Claude Code to write the mod. It reloads live

Mod vs hook vs skill

• Skill = instructions Claude reads
• Hook = a script that can allow, block or log. It can't touch the screen
• Mod = code inside Claude Code. It changes events and the screen

For teams

• Team and Enterprise load a "sec-default" guard first, so installed mods can't override your deny rules

The honest part

• Not sandboxed. A mod can read your files and secrets and act as you
• Chat to chat passes a message, not the full history
• CLI v2.1.287+ and the desktop Code tab. On by default

My take

Hooks let us react. Mods let us reshape. My first mod: an "are you sure?" before any command touches prod.

What would your first mod do? 👇

Sources

• Anthropic, Customize Claude Code with mods (1 Oct 2026)
• Claude Code Docs, Mods overview