Claude Code Mods: What They Are and the Best Ones to Install

Claude Code mods are plugins that run code inside Claude Code: they can draw panes, hold tool calls and add commands. They became public in Claude Code v2.1.287, where the changelog says "Added Claude Mods: plugins may now modify deeper behavior". Anthropic announced them on October 1, 2026 in Customize Claude Code with mods. Below: how mods differ from hooks and skills, how to install one, and the ten we would try first.

What are Claude Code mods

The mods overview in the Claude Code docs defines a mod as "a plugin that changes how Claude Code looks and behaves." A mod is a set of JavaScript or TypeScript event handlers. Claude Code calls a handler when something happens, such as a tool call, a submitted prompt or a part of the interface being drawn. The handler can watch the event, change it, or take it over.

According to the docs, a mod can:

• draw a pane beside the transcript or a band above the prompt, with buttons and text fields
• redraw parts of Claude Code's own interface, such as a tool call's row or the spinner
• hold a tool call to ask you first, or answer it without running the tool
• add a /command that runs your function at once, with no Claude turn

Mods work in the Claude Code command-line tool, the Desktop app's Code tab and the VS Code extension. In the terminal you need v2.1.287 or later, and mods are on by default. The early access variable CLAUDE_CODE_ENABLE_FUNCTION_HOOKS is now ignored.

Some of Claude Code's own features are mods, such as cc-plugin-diff, which draws the /diff pane. In a GitHub issue about the account email in the User-Agent header, a Claude Code maintainer replied that a new built-in mod, cc-plugin-privacy, blocks this from v2.1.317. One later comment reports a case it missed on v2.1.319. The mod's source is public.

Mods vs plugins, skills and hooks

Mods are plugins. They "ship inside plugins," as the announcement puts it, so you install and share them like any plugin. A small mod is three files: a plugin.json manifest, a hooks/hooks.json that points to the code, and the code file.

The difference is where the code runs. This table is shortened from the one in the docs:

	Mod	Settings hook	Skill
What it is	Functions Claude Code calls in its own process	A shell command, HTTP request or prompt run on a lifecycle event	A SKILL.md file of instructions Claude reads
What it can change	Tool calls, prompts, commands, turns and what the interface draws	Whether a tool call or prompt goes ahead, and a tool call's arguments and result	What Claude knows and does
Can it draw in the interface	Yes	No	No
What you write	JavaScript or TypeScript	A script and a settings.json entry	Markdown

The docs ask you to compare these before you write a mod: a settings hook is enough to block or log an event, and a skill is enough when you keep pasting the same instructions.

How to install a mod

A mod installs as a plugin from a marketplace. Add the marketplace once, then install the plugin by its name, an @, and the marketplace's name. This example installs nextcost.

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add davepoon/buildwithclaude
   

2. Inside Claude Code, install the mod. This opens the plugin's details so you can read what it adds and pick a scope:

   /plugin install nextcost@buildwithclaude
   

3. If the install summary says Run /reload-plugins to apply., type this inside Claude Code:

   /reload-plugins
   

4. Inside Claude Code, check what loaded. A dim line under the tabs shows a count and names, such as 1 mod active · first-mod:

   /plugin
   

To do the same install from your shell instead:

1. In your terminal, add the marketplace:

   claude plugin marketplace add davepoon/buildwithclaude
   

2. In your terminal, install the mod:

   claude plugin install nextcost@buildwithclaude
   

3. In a Claude Code session that is already open, load it (new sessions load it on their own):

   /reload-plugins
   

To start one session with no installed mods:

1. In your terminal, run:

   claude --safe-mode
   

These steps come from the install docs and the mods overview.

Mods are not sandboxed

The announcement says mods "aren't sandboxed" and run with the same machine access as Claude Code. Per the docs, a mod can read and write your files, make network requests, read API keys in your environment, see every prompt and tool call, and approve a tool call before you're asked. Claude Code's sandbox covers only the Bash commands Claude runs. A process that a mod starts runs outside it.

So read the source of a mod before you install it, and prefer marketplaces and authors you already trust. Claude Code can list what a mod does without running it.

1. Clone the mod's repo to your machine.

2. In your terminal, run this (replace ./some-mod with the path to the cloned plugin):

   claude plugin validate ./some-mod
   

The hooks: and calls: lines show the events the mod handles and what it asks Claude Code to do, such as reading a file or making a network request. For the wider picture on plugins, MCP servers and permissions, see our post on Claude Code security.

Best Claude Code mods in our catalog

As of November 14, 2026, 112 plugin listings in our catalog match our mod filter (the word mod, or a pane or band above the prompt). Five are false matches, such as a library catalog tool that reads MODS metadata. That leaves 107 mods in 48 marketplaces.

We picked ten. Our rules: the description says plainly what the mod does, the repo had commits in November 2026, and each pick does a different job. We did not rank them by stars, since the GitHub stars on a marketplace count the whole repo.

Two to start with

atc-blast-radius holds risky shell commands such as rm -rf, git reset --hard, git clean, force pushes and migrations. It shows what they would delete, and you choose Proceed or Cancel. Anthropic ships a sample mod with the same idea, blast-radius, in its playground repo.

nextcost is a band above the prompt that prices the next message before you send it. It counts down to the prompt cache going cold and shows what re-warming it will cost, with session cost, context fill and the five-hour and weekly limits beside it. Its description says it never calls a model or sends data.

Eight more

Mod	What it does
secrets-veil	Masks secret values in tool results before the model reads them
clawd-band	An animated Clawd sprite with five-hour and weekly usage bars and a context bar; the sprite needs kitty's graphics protocol
pinboard	Pins replies, notes and checklists in a side pane with /pin; pins without a title get one from Haiku, through your own session
atc-filetree	Your folder's files in a pane, with git status and the files Claude reads, writes and commits lit up
crit-tui	A review pane for the conversation, your diff or a file, with Claude replying in threads; needs the crit command-line tool
agent-radar	One live line per running subagent or background shell: model, time, tool count, current task
subagent-tuner	Lets you choose the model and reasoning effort of every subagent, or save a default
harness-scope	Turns global skills, agents, instruction files and tools on or off per repo through named profiles; it only reads, by its author's account

Install commands by marketplace

Type these inside Claude Code. For each marketplace, add it once, then install only the picks you want from it.

buildwithclaude

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add davepoon/buildwithclaude
   

2. Inside Claude Code, install nextcost:

   /plugin install nextcost@buildwithclaude
   

3. Inside Claude Code, install clawd-band:

   /plugin install clawd-band@buildwithclaude
   

4. Inside Claude Code, install pinboard:

   /plugin install pinboard@buildwithclaude
   

5. Inside Claude Code, install subagent-tuner:

   /plugin install subagent-tuner@buildwithclaude
   

6. Inside Claude Code, install harness-scope:

   /plugin install harness-scope@buildwithclaude
   

atc-agentic-toolkit

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add atc-net/atc-agentic-toolkit
   

2. Inside Claude Code, install atc-blast-radius:

   /plugin install atc-blast-radius@atc-agentic-toolkit
   

3. Inside Claude Code, install atc-filetree:

   /plugin install atc-filetree@atc-agentic-toolkit
   

orchestkit

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add yonatangross/orchestkit
   

2. Inside Claude Code, install secrets-veil:

   /plugin install secrets-veil@orchestkit
   

crit

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add tomasz-tomczyk/crit
   

2. Inside Claude Code, install crit-tui:

   /plugin install crit-tui@crit
   

claude-code-mods

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add hamzafer/claude-code-mods
   

2. Inside Claude Code, install agent-radar:

   /plugin install agent-radar@claude-code-mods
   

What we would install first

Start with a guard. atc-blast-radius is the one we would add first, because it stops rm -rf and force pushes before they run. Add secrets-veil next if your tool output often holds keys. Then nextcost, which shows what the next message will cost while you can still change it. We would leave panes for later.

Install one mod at a time. Each one sees your whole session.

To build your own, the built-in plugin-authoring skill lets you describe a mod in a session and have Claude write it, as the create a mod guide explains. modsmith is a skill for the same work: ready-made mod patterns, a review of per-turn token and prompt-cache cost, and steps for vetting someone else's mod. To install it:

1. Inside Claude Code, add the marketplace:

   /plugin marketplace add DannyMac180/skills
   

2. Inside Claude Code, install modsmith:

   /plugin install modsmith@modsmith