TryBrake for Claude Code
Let Claude Code work. Never let it run rm -rf.
TryBrake hooks into Claude Code and checks every shell command and file edit before it happens. Dangerous ones are blocked, sensitive ones wait for you, and everything is logged.
The same install works for Codex CLI, Cursor and Gemini CLI.
How it works
Set up in three steps.
- 1
Copy the install text
In TryBrake, create an agent for Claude Code. You get a short text to paste.
- 2
Paste it into Claude Code
Claude Code runs a one-time installer that adds TryBrake's hook to your project settings, keeping a backup of the old file.
- 3
Code as usual
Each command and edit is checked against your rules in a fraction of a second. You see everything in the dashboard.
What it stops
The mistakes you can't undo.
Folder and disk wipes
rm -rf and its variants are recognised however they're written, and blocked.
Database wipes
DROP TABLE, DROP DATABASE and TRUNCATE are blocked.
Secrets
Edits to .env files are blocked, so keys don't get overwritten or leaked.
Risky git
Force pushes and history rewrites wait for your approval.
Questions
Good to know.
Does it slow Claude Code down?
Barely. Each check is a single quick request, and only commands and file edits are checked, not reading files.
What if TryBrake is unreachable?
The command is blocked and Claude Code is told why. Nothing risky runs without a check.
Can I remove it?
Yes. The installer keeps a backup of your settings file, and the hook is a few lines you can delete.
Works with more than Claude Code.
One TryBrake account protects every agent you run, wherever it's built.
- n8nProtect workflows and AI Agent tools
- MakeGuard modules in your scenarios
- ChatGPT & OpenAICodex CLI and agents built on OpenAI
- ZapierOne webhook step before the action
- CursorTerminal commands and MCP tool calls
- Gemini CLICommands and file edits
- PythonLangChain, CrewAI, OpenAI Agents SDK…
- JavaScript / TypeScriptVercel AI SDK, LangChain.js, Node
- Anything elseOne HTTP call before each action

