Anthropic added mods to Claude Code on 1 October 2026. A mod is a small piece of JavaScript or TypeScript that changes how Claude Code looks and behaves, and you can ask Claude to write one for you. This guide gives you three ready prompts: sessions that talk to each other, a usage watcher with a cheaper "lite mode", and a personal dashboard.
What you need
Claude Code, the coding tool, not the regular Claude chat app. Mods work in the terminal version, the Code tab of the Claude Desktop app and the VS Code extension. See the mods overview.
A paid Claude plan. Claude Code is not on the Free plan. Pro costs $20 a month, or $17 a month billed yearly (Claude pricing).
Claude Code v2.1.287 or later in the terminal, or v2.1.286 or later inside the Desktop app.
Step by step
Check your version. In your terminal, run the command below. If it is older than 2.1.287, update Claude Code first.
claude --versionOptional: try an official sample mod. Anthropic shares sample mods in its claude-code-playground repo.
token-weatherdraws a forecast of your context window above the prompt. Clone the repo, then load the mod for one session:
claude --plugin-dir claude-code-playground/claude-code/mods/token-weatherOpen a Claude Code session in any folder and paste one of the prompts below. Claude uses a built-in skill called
plugin-authoringto write the mod. When it saves the first file, Claude Code asks whether to enable hot reloading. Choose Enable for this session. The mod loads when Claude finishes its turn.
Mod 1: sessions that talk to each other
You may not need a mod for this. Claude Code already lets one session message another on the same machine (cross-session messaging). Just type something like "Let @research know the list is ready". A mod adds your own shortcut and a log of what arrives:
Make a mod called team-board. Add a /tell command that takes a session name or id and a message, and sends the message as plain text to that session using $.session.send. If it is not delivered, show a toast with the reason. Also add a pane, opened with /team, that logs every message this session receives (session.receive event) with the sender and time, and passes each message through unchanged. Do not approve tool calls or change any permissions.A good split is one session that researches, one that drafts, and one that reviews as the "manager". You still read and approve anything that leaves your computer.
Mod 2: usage watcher with a lite mode
A mod can read the token use of each request and can send a request to a different model. This one shows your usage and lets you switch to a smaller model for simple work:
Make a mod called cost-watch. After each model request (turn.step), read the token usage and show a running total for this session in the band above the prompt: input tokens, output tokens and cache reads. Add a /lite command that toggles lite mode. When lite mode is on, send main-conversation requests (not subagent requests) to the haiku model with next({ ...e, model }). When it is off, leave the model unchanged. Show "LITE" in the band while it is on. Save the toggle with $.store so it survives a restart.Use lite mode for renaming, small edits and quick questions. Turn it off for planning and hard bugs.
Mod 3: your own dashboard
Mods can draw a pane beside the conversation with tabs, buttons and text fields:
Make a mod called my-desk. Add a /desk command that opens a pane beside the transcript with three tabs. To-do: a text field to add items and a list with a Done button on each item. Notes: one text field that saves as I type. Repo: the current git branch and the output of git status --short, refreshed every 60 seconds. Save the to-dos and notes with $.store so they stay between sessions. Where a pane can't be drawn, have /desk print the same information as text.Check it loaded. Run
/pluginand open the Installed tab. Your mod is listed there, and you can turn it off from the same screen.Keep it. A mod Claude writes lives in
~/.claude/dev-mods/under that session's ID and is cleaned up later. Copy its folder somewhere safe, such as~/mods/my-desk, and start future sessions with:
claude --plugin-dir ~/mods/my-deskBefore installing anyone else's mod, inspect it. This lists what it hooks and what it can do, without running it:
claude plugin validate ./some-modGood to know
Mods are not sandboxed. A mod runs with your permissions. It can read files and API keys, see your prompts, and approve tool calls. Only install mods from people you trust.
Lite mode trades quality for usage. Smaller models make more mistakes. Every request a mod sends also counts against your plan or API bill.
A "messages to answer" panel needs your inbox. That means connecting email or chat to Claude Code. Think about privacy before you do.
Some places are limited. Plugins do not load in WSL sessions of the Desktop app. On Team and Enterprise plans, an admin may block mods.
The API is new. Events and method names can change between releases. Claude reads the type files for your version, so ask it to fix any errors after an update.
Sources
Found this useful? Follow @kevinsingh.me on Instagram for a new AI tool every day, and subscribe so you don't miss the next one.