“No reason why everyone should have an identical Claude experience”: Anthropic’s mods let you change Claude Code’s look and behavior

Anthropic is giving Claude Code a much deeper customization layer. The new feature, called mods, lets developers hook JavaScript or TypeScript functions into events inside the product, so they can change what Claude sees, what it does, and even parts of the interface itself.

That’s a step beyond the usual knobs. Developers already had settings, CLAUDE.md instructions, hooks, MCP servers, status lines, and output styles. Those could shape prompts, permissions, and tools. Mods go further. A mod can rewrite a prompt before it reaches the model, intercept or alter a tool call, handle permission prompts, replace pieces of the UI, or add entirely new behavior.

Claude Code creator Boris Cherny framed it as a rejection of one-size-fits-all software: people work differently, so there’s no reason everyone should have the same Claude experience. He said mods can be prompted in a simple way and packaged as plugins that other users can install. Anthropic is already using the system internally for features like AGENTS.md support and the /diff pane, and has published that source in the Claude Code repository.

The launch also makes Claude Code feel more alive while it runs. A mod can keep state during a session, so it can remember information between events, update displays in real time, launch processes, add commands, or expose new tools to the model. In one example, Addy Osmani of Anthropic showed a roughly 80-line mod called Token Weather, which tracks context usage and changes its label from “Clear” at 18% of a 200,000-token window to “Showers” at 67% and “Storm” at 81%.

Mods are enabled by default in Claude Code 2.1.287 and later, and they’re available now in the CLI and desktop app. But there’s a catch, because there almost always is. Since mod code runs locally on the user’s machine, Anthropic says to inspect the source and only install code from trusted publishers. Team and Enterprise users, plus machines with managed settings, get Anthropic’s sec-default guard ahead of user-installed mods, which blocks those mods from overriding permission-deny rules by default.