# Claude Code Mods let you change the interface and how it works

> Anthropic lets developers customize Claude Code with mods. Explore three official demos and learn how mods differ from plugins and MCP.


Published: 2026-10-02. Original source: [Anthropic · Customize Claude Code with mods](https://claude.com/blog/claude-code-mods).

**Anthropic introduced Claude Code Mods on October 1.** Developers can add interface elements, change how tools run and replace assistant features with small TypeScript modules. Mods work in the Claude Code terminal and desktop app.

With mods, a team can adapt its coding session without waiting for an Anthropic update. The [official announcement](https://claude.com/blog/claude-code-mods) explains how to install and share those modifications. The cover illustration is also by Anthropic.

## Three demos that show what changes

The [Anthropic tutorial](https://claude.dev/blog/getting-started-with-claude-code-mods/) presents three examples. Select one and play the recording; switch to full screen to read the terminal more closely.

**Token Weather** displays context window usage: how much space the conversation takes up in the material Claude uses to answer. The demo moves from 18% to 67% to 81%. It lets you follow a growing session; those percentages do not measure API spending.

**Blast Radius** pauses a deletion command and lists the files it would affect before asking for confirmation. Its text-based classification can miss commands hidden in aliases or scripts. Permission rules remain necessary.

**Replay Theater** lets you step through a turn's file edits and review their diffs. The view helps you inspect the sequence of changes. The example shown does not provide an undo function.

The recordings demonstrate example features. When evaluating a mod for your project, check which actions it intercepts and which it lets through.

## What a Claude Code mod does

The [mods documentation](https://code.claude.com/docs/en/plugins/mods/overview) explains that Claude Code emits events when it receives a prompt, calls a tool or draws the interface. A mod can act on those events: observe what happens, modify it or respond in its place.

For example, a mod can rewrite a prompt before it reaches the model, block a call or add a pane. Anthropic also lets you ask Claude to build the mod and reload it during a session. The tutorial requires **Claude Code 2.1.287 or later** and notes that the API can change between releases.

Some Claude Code features already use this structure. The `/diff` pane, which lets you review file differences, is now a mod that you can disable or replace. A team building its own review view should check which changes it shows before using it to approve work.

## Mods, plugins and MCP: what each does

Mods ship inside [plugins](https://code.claude.com/docs/en/plugins). MCP connects the assistant to external tools and data, as described in the [Claude Code documentation](https://code.claude.com/docs/en/mcp).

| Element | What it provides | Example |
| --- | --- | --- |
| Mod | Changes Claude Code's behavior or interface. | A pane for stepping through file edits. |
| Plugin | Packages extensions you can install and share. | Distributing a mod to a team. |
| MCP | Provides authorized access to external tools and data. | Reading an account's posts and metrics. |

HeyMark's [connection to Claude](/en/guides/connect-claude-heymark/) uses MCP to let the assistant read your accounts' posts and metrics after you authorize access. You can follow that guide with the current connector, without installing a mod. HeyMark's integration with these new mods has not been tested yet.

## What to check before installing a mod

Anthropic [warns that mods have the same access to your machine as Claude Code](https://code.claude.com/docs/en/plugins/mods/overview#what-a-mod-can-reach). Review the author, code and permissions before installing one, especially if it handles tool approvals.

We suggest starting with a test copy of your project, a small task and an outcome you can check. Review the file changes and test both an action the mod should allow and one it should block. If you share it with a team, document those limits alongside the installation instructions.

You can explore the [Claude Developers thread on X](https://x.com/ClaudeDevs/status/2105721434807083061) and the [sample mods' source code](https://github.com/anthropics/claude-code-playground/tree/main/claude-code/mods) to see how they are built.

## Plan your next post in HeyMark.

Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.

[Start free](https://app.heymark.ai) · [See how it works](https://heymark.ai/en/#product)

## Developer and agent resources

- [HeyMark MCP documentation](https://heymark.ai/en/mcp/)
- [llms.txt](https://heymark.ai/llms.txt)
- [Full site content for language models](https://heymark.ai/llms-full.txt)
- MCP protocol endpoint: `POST https://mcp.heymark.ai`
- OAuth protected-resource metadata: [/.well-known/oauth-protected-resource](https://mcp.heymark.ai/.well-known/oauth-protected-resource)
- MCP server card: [/server-card](https://mcp.heymark.ai/server-card)