TryBrake for n8n
Your n8n workflows, with a brake on every risky step.
Connect your n8n and TryBrake reads your workflows, spots the steps that move money, send messages or delete data, and lets you protect each one with a switch. Nothing to rebuild.
How it works
Set up in three steps.
- 1
Connect n8n
Paste your n8n address and an n8n API key. TryBrake lists your workflows and draws each one, step by step.
- 2
Review the risky steps
TryBrake's AI reads the workflow and proposes rules for each action, like "refunds over €200 need my OK". You edit or accept them.
- 3
Switch protection on
TryBrake adds a small check in front of the step inside your workflow. Turn it off and your workflow goes back exactly as it was.
What it stops
The mistakes you can't undo.
Refunds and payments
Hold any Stripe refund above an amount for a human, or cap how many can go out per hour.
Emails and messages
Ask before your agent emails a customer, or block sending to addresses outside your company.
AI Agent tools
Each tool your n8n AI Agent can call gets its own rules. If a call is refused, the agent is told why and tells the user.
Loops and runaways
A workflow stuck in a loop is cut off by a volume limit before it sends 500 emails.
Questions
Good to know.
Does TryBrake change my workflow?
Only the steps you protect: it adds one Code node in front of each. For an AI Agent tool, the tool moves into a small helper workflow so the check runs first. Switching protection off restores the original step exactly.
What if TryBrake is unreachable?
The protected step doesn't run. Every check fails closed, so an outage can never let a risky action through.
Does it work with n8n Cloud and self-hosted n8n?
Yes. TryBrake talks to your n8n through its public API, so it works with both, as long as your n8n can be reached from the internet.
Can I try it without connecting my n8n?
Yes. You can import a ready-made workflow that already includes the check, and paste it into n8n.
Works with more than n8n.
One TryBrake account protects every agent you run, wherever it's built.
- MakeGuard modules in your scenarios
- Claude CodeBlock risky commands and edits
- 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