AI RACE— The AI Race
New Models

Anthropic Introduces Mods to Let Developers Customize Claude Code from the Inside

Anthropic has launched Mods for Claude Code, a plugin-based middleware framework that lets developers build custom commands, intercept tool calls, and modify the interface using JavaScript or TypeScript.

10/03/2026, 14:12
Anthropic ra mắt Mods cho phép lập trình viên tùy biến sâu Claude Code

Anthropic has released a new extensibility framework called "Mods" for Claude Code, opening up the internal architecture of its terminal- and desktop-based AI coding assistant. Operating essentially as in-app middleware, the plugin-based system enables developers to fundamentally modify how Claude Code operates and renders information.

Middleware Running Directly in the Tool

Mods are written as JavaScript or TypeScript functions that hook directly into core operational events. Developers can tap into user prompts, intercept tool calls, and adjust UI rendering. This design allows users to integrate custom side panels alongside the main chat, build entirely new interactive commands, or adjust how tools execute behind the scenes.

According to Anthropic, the framework is already deeply woven into Claude Code's native feature set; existing capabilities like the /diff command are built internally as Mods. To help developers get started with building their own extensions, the company has published sample Mods on GitHub.

The First Official Mod: "You Should Know"

Alongside the framework, Anthropic unveiled its first official built-in plugin named "You Should Know."

When enabled, the plugin spins up an auxiliary AI agent tasked with monitoring Claude's output in the background. If the secondary agent detects critical details, edge cases, or potential oversights that the user might have missed in the primary stream, it triggers a "Heads up" notification. Users can activate this capability directly within Claude Code using the command:

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

Security Considerations and Platform Support

Because Mods function as internal middleware, they run with the local user's full permissions and are not sandboxed. Anthropic has cautioned developers and teams to only install and run Mods originating from trusted sources. To mitigate security risks in professional settings, the system includes administrative controls allowing organizations to restrict or allow specific Mods across developer environments.

In terms of compatibility, Mods currently function across Claude Code's command-line interface (CLI) and the desktop application, with partial support available in the VS Code extension.

◗ Sources

The Decoder10/03

Related stories