product updateAnthropic

Anthropic adds Mods to Claude Code, a plugin system that hooks into tool calls, prompts and UI rendering

TL;DR

Anthropic released Mods for Claude Code, a plugin system built on JavaScript and TypeScript functions that hook into events such as tool calls, user prompts and UI rendering. Mods are not sandboxed and run with the user's permissions. They work in the CLI, the desktop app and, partly, the VS Code extension.

3 min read
0

Anthropic has released "Mods" for Claude Code, a middleware layer that runs inside the tool and lets developers change both its behavior and its interface. The feature was reported on October 3, 2026.

What Mods do

Mods are JavaScript or TypeScript functions that attach to specific events in Claude Code. The hookable events range from tool calls and user prompts to UI rendering. According to the source report, this allows developers to:

  • Add custom panels next to the chat
  • Intercept tool calls
  • Wire up entirely new commands

Anthropic says some built-in features, including the /diff command, are already implemented as Mods. Anthropic also provides sample Mods on GitHub.

Security model

Mods are not sandboxed. They run with the user's permissions, and Anthropic warns users to install them only from trusted sources. Because Mods can intercept tool calls, a malicious or poorly written Mod could affect anything the user's account can access.

Organizations can control which Mods are allowed to load. The report does not detail how that policy is configured.

Where Mods work

Surface Support
CLI Yes
Desktop app Yes
VS Code extension Partial

The report does not specify which Mod capabilities are missing from the VS Code extension.

First official Mod: "You Should Know"

The first official Mod plugin is "You Should Know." It spins up a separate agent that watches Claude's output and sends a "Heads up" message when it detects important information the user may have missed. Users enable it with:

/plugin enable cc-plugin-you-should-know@builtin

The @builtin suffix indicates it ships with Claude Code rather than coming from an external source.

What is not disclosed

This is a product feature, not a new model. No model version, context window, benchmark score or pricing change accompanies the release. The report also does not say whether Mods that run additional agents, such as "You Should Know," consume extra tokens or count against usage limits. Cost impact is not yet disclosed.

What this means

Mods move Claude Code from a fixed product toward an extensible platform. Anthropic's own decision to build core commands like /diff on the same system suggests the hook surface is substantial rather than a thin add-on, though that is Anthropic's claim and has not been independently verified.

The trade-off is security. Unsandboxed code with full user permissions, combined with the ability to intercept tool calls, creates a supply-chain risk similar to editor extensions and npm packages. The organization-level allowlist is therefore the most important control for enterprise teams, and its granularity will determine how safely Mods can be adopted at scale.

The release also puts Claude Code in more direct competition with extensible coding environments such as VS Code-based tools and open-source agents, where customization has been a selling point. A second-agent-watching-the-first pattern, as in "You Should Know," is a notable early example of what in-tool middleware enables: automated oversight of model output without changing the model.

Teams should watch for three follow-ups: documentation on the full event list, how usage of secondary agents is billed, and whether Anthropic builds a vetted distribution channel for third-party Mods.

Comments

Loading...