Menu

Can a Claude Code hook get your account banned? What the terms say

3 min read

Reading hook events from your own Claude Code is a documented use of a documented feature, and nothing in Anthropic’s published pages forbids it. The risk starts when a tool handles your Claude sign-in or drives Claude Code for you.

This is a reading of three public pages as they stood on 11 Nov 2026. It is not legal advice, and only Anthropic decides what happens to an account.

What a hook is

A hook is a command or a web request that Claude Code runs on your own machine when something happens in a session: a prompt is sent, a file is about to be edited, an edit has finished, the session stops. You add hooks yourself, in your own settings file.

The hooks reference shows hooks posting their event to a local web service. It describes notification hooks as “intended for side effects such as forwarding the notification to an external service”. Hooks were made to tell other programs what Claude Code is doing.

What the pages say

Three pages matter.

The hooks reference. It sets no limit on what the program that receives a hook may do with the event.

The Claude Code legal and compliance page. It covers credentials. A third party may not offer Claude sign-in or route requests through your plan’s credentials. It also says those rules do not stop you signing in to the unmodified Claude Code with your own subscription.

The Consumer Terms. They cover Free, Pro, Max and Ultra accounts. They forbid accessing the service “through automated or non-human means, whether through a bot, script, or otherwise”.

Read together, they draw two lines.

The two lines

Line one: your sign-in. A tool that asks for your Claude login, stores your credentials or sends requests through your subscription is on the wrong side. So is a tool that ships its own changed copy of Claude Code.

Line two: who is typing. You are meant to be the one using Claude Code. A tool that starts sessions for you, feeds them prompts from a script or keeps them running while you sleep is doing what the Consumer Terms describe.

A tool that only listens is on the safe side of both. It never sees your sign-in, and it never sends Claude Code anything.

Three questions for any tool

Ask these before you install anything that works with Claude Code.

  1. Does it ever ask for my Claude login or an Anthropic key for my own sessions? If yes, find out exactly where that credential goes.
  2. Does it start or steer Claude Code, or do I? Look for anything that launches sessions or sends prompts on your behalf.
  3. What does it read, and where does that go? Hook events include your prompt and the path to your transcript. A careful tool tells you which fields leave your machine and which never do.

If a tool’s page does not answer these, ask its maker. A clear answer takes one line.

How Flockshow answers them

Flockshow shows each Claude Code session as a named cursor on your running app, so it is fair to hold it to the same three questions.

  1. Sign-in. It never asks for your Claude login and never handles a credential. You sign in to Flockshow itself with GitHub or Google.
  2. Driving. It never launches Claude Code and never sends it a prompt. You type in your own terminal. It listens to the hook events and draws cursors.
  3. What it reads. It reads which file a session is editing and the page text that will change. Prompts, paths and source code stay on your computer.

Its name does not use “Claude”, and Anthropic does not make or endorse it.

What nobody can promise

No tool maker can promise what Anthropic will do, and terms change. If your work depends on your account, read the three pages yourself. They are short.