Claude Code adds mods that change how it looks and behaves

Anthropic has added mods to its coding agent Claude Code. Shipped with version 2.1.287 on October 1, 2026, mods are small plugins that change how Claude Code looks and behaves. Developers can use them to add their own panes, commands and tool call rules to Claude Code.

What mods can do

According to the Claude Code docs, a mod is a plugin made of event handlers written in JavaScript or TypeScript. When Claude Code is about to call a tool, handle a prompt you submitted or draw part of the screen, it calls the relevant mod, which can watch the event, change it or take it over. That lets a mod draw a pane with tabs and buttons beside the transcript, restyle parts of Claude Code's own interface, hold a tool call while it asks you a question, answer a tool call without running the tool, or send a request to a different model. A mod can also add custom commands that run at once, without waiting for Claude.

Anthropic built some of Claude Code's own features as mods, including the /diff pane and the feature that loads AGENTS.md as project instructions; the source of some of these built-in mods is public in the Claude Code repository. A built-in mod called "You should know", added in the same release, runs a side agent during longer tasks and shows a note above the prompt when it finds something you or Claude might miss. It is off by default.

Getting and turning off mods

Mods are on by default in Claude Code 2.1.287 and later. They work fully in the CLI and in the Code tab of the Claude desktop app; in the VS Code extension and with claude -p, a mod's code runs but what it draws does not appear. The easiest way to get one is to describe what you want in a session and let Claude write the mod. A ready-made mod installs from a plugin marketplace with /plugin install. Anthropic also shares sample mods in a separate repository, such as one that shows how full the context window is and one that holds risky commands like rm -rf for confirmation.

A single mod can be turned off from the /plugin screen. Starting Claude Code with --safe-mode stops installed mods for that session, and setting "disableAllHooks": true stops them in every session. In organizations, administrators can limit which mods load through managed settings.

The security side

Anthropic's docs state that mods are not sandboxed. A mod runs with your permissions: it can read and write files anywhere your user account can, start programs, make network requests, read API keys kept in environment variables or settings files, see every prompt you send and every tool call Claude makes, approve a tool call before you are asked, and spend model usage on your plan. Even with sandboxing turned on for Bash commands, processes a mod starts run outside the sandbox. For that reason, the company advises installing mods only from authors and marketplaces you trust. Before installing one, claude plugin validate lists the events it handles and what it asks Claude Code to do, without running it. According to the docs, mods cannot restyle the permission prompt.

Mods open up a lot of room to customize Claude Code, and every mod you install is code that can act on your machine as you. It is wise to skip mods whose source you have not seen, to list what a mod does before installing it, and to stay within the limits your administrator sets on work machines. To see how the feature holds up in practice, it is worth following evaluations from independent researchers as well.

Sources: Anthropic: Claude Code v2.1.287 release notes, Claude Code docs: Mods overview, Claude Code docs: Manage mods for your organization.