Launching today

TryBrake
A kill switch and human approval for your AI agents
2 followers
A kill switch and human approval for your AI agents
2 followers
TryBrake sits between your AI agents and the actions they take. Risky ones (refunds, emails, deletes) are held for a human to approve, rules block the rest, and one switch stops every agent. Works with n8n, Make, Zapier, Claude Code. 14 days free.








Free Options
Launch Team

SavrdA journal you talk to. Voice AI writes the entry.
Promoted
Hi Product Hunt, I'm Élie, the maker of TryBrake.
AI agents can already refund customers, send emails and delete records. Most teams I talk to have nothing standing between the agent and the action. The real examples are public: an AI agent deleted a live production database during a code freeze, and an airline was held liable for what its chatbot promised.
TryBrake is that missing layer:
• Risky actions are held until a human approves them (one click)
• Rules decide what's allowed, what's blocked, and what needs a human
• One switch stops every agent at once
• Works with n8n, Make, Zapier and Claude Code
There's a free 14-day trial (no credit card), and a free exposure check on the homepage: paste your company website and see what an unguarded agent could do.
What I'd love to hear from you: what's the riskiest thing your agent can do on its own today?
@trybrake_saas How is this different from n8n’s built-in approval?
@efer_zaetz Good question! n8n's approval is per workflow: you set it up in each flow, and it lives inside n8n. TryBrake sits outside, in front of the action itself, so the same rules, one audit trail and one kill switch work across n8n, Make, Zapier, Claude Code and your own code. If you only use n8n and a few flows, the built-in step may be enough. If you run several tools or agents, TryBrake gives you one place to control them all. Happy to show you a quick demo if useful!