Editorial illustration for Claude Code's New Mods System Lets Developers Rewrite the AI Tool Internally
Claude Code Mods Let Developers Customize AI Tool
Anthropic shipped a feature called Mods for Claude Code this week, and it changes the tool from something you run to something you can rebuild. Mods are JavaScript or TypeScript functions that hook into events inside Claude Code itself, tool calls, user prompts, UI rendering, giving developers a way to reach into the application and change how it behaves rather than just what it outputs.
The company says some of Claude Code's existing features, including the /diff command, are already built using this same system, which suggests Mods aren't a bolted-on extra but closer to the actual architecture the tool runs on. Anthropic has posted sample Mods on GitHub for developers who want to start building their own.
There's a catch worth noting before anyone starts installing things: Mods run with full user permissions and carry no sandboxing. Anthropic is direct about the risk, telling users to only grab Mods from sources they trust. Organizations get a say too, with controls to restrict which Mods are allowed to load. Support spans the CLI and desktop app, with partial coverage in the VS Code extension.
The plugin-based system lets developers change how Claude Code looks and works. Mods are JavaScript or TypeScript functions that hook into specific events, from tool calls and user prompts to UI rendering. That means developers can add custom panels next to the chat, intercept tool calls, or wire up entirely new commands.
Why this matters Anthropic just handed developers a scalpel for Claude Code's internals, and that's a bigger deal than another feature release. Letting JavaScript and TypeScript functions hook into tool calls, prompts, and UI rendering turns Claude Code from a fixed product into a platform, closer to how VS Code extensions reshaped that editor than to a typical AI tool update. The fact that /diff is already built as a Mod suggests Anthropic trusts this system enough to run its own features through it, which is a real signal of commitment, not a side experiment.
For founders, this opens a path to build proprietary workflows directly inside Claude Code instead of wrapping it with external scripts. For developers, it means auditing what third-party Mods actually touch, since intercepting tool calls is also a way to intercept sensitive data. We'd watch whether Anthropic publishes stable APIs for these hooks or lets them shift release to release, because that decision determines whether a Mods ecosystem actually forms or stays a niche for a few motivated tinkerers.
Related Reading
Browse all Anthropic news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic Claude gains shared context in Excel, PowerPoint; Microsoft adds Copilot Cowork
AI Tools & Apps
Related Topic Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source
From Archives FedCausalCompose Framework Links LLM Agent Actions to World Models
From Archives Anthropic co-founder fears creating AI that "suffers perpetually
From Archives Pentagon Labels Anthropic a Supply Chain Risk Over Weapons Ban
From Archives Google Ditches Gems, Adopts Anthropic's Open-Source Skills Format
Open Source
Common Questions Answered
What are Mods in Claude Code and how do they work?
Mods are JavaScript or TypeScript functions that hook into specific events within Claude Code, including tool calls, user prompts, and UI rendering. This allows developers to reach into the application's internals and modify how it behaves rather than just changing its outputs, essentially turning Claude Code from a fixed product into a customizable platform.
What can developers do with Claude Code Mods?
Developers can use Mods to add custom panels next to the chat interface, intercept and modify tool calls, wire up entirely new commands, and change how Claude Code looks and works. The system provides granular access to the application's internal events, enabling deep customization beyond typical plugin functionality.
How does the Mods system compare to VS Code extensions?
Similar to how VS Code extensions reshaped that editor by allowing developers to extend its functionality, Anthropic's Mods system transforms Claude Code from a fixed product into a platform that can be fundamentally modified and extended. This represents a significant shift in how AI tools can be customized compared to traditional feature releases.
Is the Mods system already being used for Claude Code's existing features?
Yes, Anthropic has already built some of Claude Code's existing features, including the /diff command, using the Mods system itself. This demonstrates that Anthropic trusts the Mods architecture enough to run its own core features through it, validating the system's reliability and capability.
Further Reading
- Customize Claude Code with mods in TypeScript - Claude by Anthropic
- Claude Code Gets Mods, Plugins That Rewrite the Agent - Pasquale Pillitteri
- Anthropic lets Claude Code mods rewrite prompts and approve tool calls - runtimewire.com
- Anthropic Launches Claude Code Mods, TypeScript Agent Hooks - AI Weekly
- Claude Code gets Mods. A 5-minute guide to the system that lets you customize AI agents like game mods - note.com
