Claude Mods Guide

Mods are small JavaScript or TypeScript functions that change how Claude Code looks and behaves. Anthropic launched them on October 1, 2026. They ship inside plugins, so you install and share them like any plugin.

Claude Code fires an event whenever it does something, such as calling a tool, receiving a prompt, or drawing part of the screen. A mod hooks into an event and can:

Observe

Watch the event and let it continue

Rewrite

Change the event before it continues

Answer

Handle it itself, such as refusing a risky command

Mods let you shape Claude Code around how you work.

• Draw a pane beside the transcript or a band above the prompt, with buttons and text fields
• Restyle parts of Claude Code’s own interface, such as the spinner or a tool call row
• Hold, rewrite, or block a tool call
• Rewrite a prompt before it reaches the model
• Add a /command that runs your code instantly, without a Claude turn
• Hide secrets from tool output before Claude reads it

A mod cannot change what the permission prompt shows you.

Terminal

Yes, with Claude Code v2.1.287 or later

Desktop app Code tab

Yes, from v2.1.286, except some terminal-only elements

VS Code chat panel, claude -p, and cloud sessions

A mod’s logic runs, but what it draws does not appear

Mods are on by default.

Some Claude Code features are now mods. See them in /plugin, under Installed, then Built-in:

/diff

The diff pane. You can turn it off or replace it with your own.

AGENTS.md

Loads AGENTS.md files as project instructions

You Should Know

Off by default. A side agent watches longer tasks and shows a note above the prompt when it spots something you might miss.

Security default

Protects what your organization manages from user-installed mods

Anthropic plans to move more built-in features to mods over time.

• In Claude Code, run /plugin install name@marketplace
• Or in your shell, run claude plugin install name@marketplace
• Browse plugins in the Claude directory or from marketplaces you trust
• Run /reload-plugins to load it in an open session
• Run /plugin to see which mods are active

Anthropic shares sample mods in its claude-code-playground repository on GitHub:

Token Weather

A forecast of how full your context window is, above the prompt

Blast Radius

Holds risky commands, such as rm -rf or a force push, and shows Proceed or Cancel buttons

Replay Theater

Adds /replay to step through the file edits Claude made in the last turn

Load one for a single session with claude --plugin-dir and the mod’s folder.

Describe the mod you want in a Claude Code session, and Claude can write it, load it, and reload it as you change it. You can also write the code yourself.

Make a mod that shows how many tool calls you have made this turn beside the spinner.

Mod

Code inside Claude Code. Use it for panes, commands, and rewriting events.

Settings hook

A script that runs on an event. Use it to block, allow, or log with a script you already have.

Skill

Written instructions. Use it for instructions you keep repeating.

MCP server

Gives Claude new tools. Use it to reach outside systems.

Pick the simplest one that does the job.

One mod

Disable or uninstall its plugin in /plugin

All mods, one session

Start Claude Code with --safe-mode

All installed mods, always

Set disableAllHooks to true in your settings. This also stops your settings hooks and custom status line.

• Mods are not sandboxed and run with your permissions
• A mod can read your files and secrets, see your prompts, approve tool calls, and use your plan
• Install only from authors and marketplaces you trust
• Run claude plugin validate on a mod’s folder to list what it does before installing
• Read the code of any mod you install

A mod is code with your keys. Treat it like any software you install.

Admins control mods through managed settings and plugin marketplace rules. On Team and Enterprise plans, and on managed machines, the built-in security mod loads first and limits risky behavior from user-installed mods.