Guide · Claude Code mods

How to use Claude Code mods like the top 1%

One mod sends every subagent to the cheapest Claude that can do its task, and shows you the bill. The other opens a Tetris while Claude works. Both are one prompt each, plus 20 more from the mods people shipped in week one.

26 steps · ~10 min to your first modClaude Code v2.1.287+ · no paid third-party serviceHaiku 4.5 · Sonnet 5.5 · Opus 5.5 · Fable 5.1

What you will build

Two mods, each from one prompt you paste as is.

The model router: 4 roles (architect plans, builder writes the code, runner runs the tests, fixer fixes what fails). Haiku reads every task first and labels it simple, standard, hard or long run. The mod starts the subagent on Haiku 4.5, Sonnet 5.5, Opus 5.5 or Fable 5.1. A side pane shows what each agent cost, and what the same work would have cost on Fable.

The Tetris: it opens by itself when Claude is still working 2 seconds after your prompt, and pauses and gets out of the way when Claude is done.

Then a bonus: 20 real mods people shipped in the first week, sorted by what they are for, each credited to its author and each with a prompt that builds your own version.

Prerequisites

What you need before you start

01 · Claude Code v2.1.287 or later

Mods launched in this version. Check with claude --version.

02 · A wide terminal

144 columns or more, so a pane a mod opens by itself can appear.

03 · Nothing else

No paid third-party service. Every prompt runs on Claude Code alone.

Unlock the full guide

Enter your email to read the rest of the tutorial.

Copy-paste code. Any email works. Unlocks in your browser, so you only do this once.

Step 01

What a mod is, in five lines

Anthropic launched mods on 2026-10-01 in Claude Code v2.1.287 (announcement, docs).
A mod is a plugin whose JavaScript or TypeScript handlers run inside Claude Code on events: a tool call, a prompt, a model request, a part of the UI being drawn. A handler can watch the event, rewrite it, or answer it itself.
It can draw a pane beside the transcript or a band above the prompt, restyle the spinner and the tool rows, hold or rewrite a tool call, send a request or a subagent to another model, and add /commands.
Install one with /plugin install <name>@<marketplace>, or for one session with claude --plugin-dir <folder>. Check one before you install it: claude plugin validate <folder>. To write one, ask Claude: it uses its built-in plugin-authoring skill.
A mod runs with your permissions and is not sandboxed. It can read your files and keys and approve tool calls. It cannot change the permission prompt. claude --safe-mode, or "disableAllHooks": true in your settings, turns every installed mod off.

Step 02

How to use the prompts in this guide

Every prompt is complete. Paste it as is, no assembling. They were written against the mod API types of Claude Code 2.1.292, and every event and method they name exists in those types.

Update Claude Code to 2.1.287 or later (claude --version).
Open Claude Code in any folder and paste one prompt. Claude loads plugin-authoring, writes the mod in its own mods folder, validates it and tests it.
Claude Code asks once: "Enable hot reloading for this session?". Answer Enable for this session. The mod loads when Claude's turn ends and reloads on every later edit.
Follow the 3-step live test Claude gives you at the end. If something is off, describe what you see in one sentence. Claude edits, the mod reloads.
To keep a mod in every session, see Step 05.

Unlock above to read the full guide

Want us to build agents like this for you?

Want this built for you?

Mods are one way to make Claude Code fit how you work. ColdIQ builds AI-powered GTM systems for B2B teams every day: the data layer, the agents on top of it, and the sequences that turn the output into conversations.

Talk to us

Your GTM stack. One API key.

Stop managing 39 tools. Start shipping GTM from code.

Get Started Or try the playground first

Free to start. No credit card required.