Security
Built to sit in the middle of your most sensitive actions.
TryBrake decides whether an agent may refund, email or delete. Here is how we make sure that decision is safe, and that your data is too.
Decisions without AI
Your rules are plain logic, checked in order before every action: allow, block, or ask a human. No AI takes part in the decision, so no prompt, however clever, can talk its way past a rule.
Fails closed
Every connector blocks the action when TryBrake can't be reached or answers with an error. An outage, a timeout or a bad network never lets a risky action through.
We don't hold your agents' keys
Your agent keeps its own access to Stripe, Gmail or your database. TryBrake only answers yes, no or wait, so a breach on our side can't move your money or read your mail.
One-click kill-switch
Stop every agent of a workspace at once, stop a single agent, or switch off one action. It takes effect on the very next call.
Data and hosting
- Database and sign-in hosted in the EU, in Paris, France.
- Every workspace is walled off from the others by row-level security in the database itself, not only in the app.
- History is deleted automatically when it passes your plan's retention: 30 days on Essential, 1 year on Pro.
- All traffic is encrypted in transit with HTTPS.
Keys and secrets
- Agent API keys are stored as a SHA-256 hash: we can check a key, but we can't read it back.
- If you connect n8n or Make, that API key is encrypted with AES-256-GCM and used only to add or remove protections in your workflows.
- Install links for coding assistants work once, expire after 15 minutes, and never contain the key itself.
Approvals
- Approval links sent to Slack, Teams, Google Chat or Telegram are signed, work for one action only, and expire after 12 hours.
- Opening a link decides nothing: approving or denying takes a click, so link previews and security scanners can't approve by accident.
- Each decision is recorded with who made it and when.
Accountability
- Every action an agent asks for is logged with its details, the decision and the reason.
- Every change to rules, limits, agents and settings goes into an audit log you can export, with a printable report.
- Rules have a version history, so you can see what changed and when.
Team access
Invite your team as members or admins. Only owners and admins can change rules, keys and billing.
AI, only to suggest
AI (Anthropic's Claude) helps write rules and analyse workflows. It only suggests: you review every rule, and decisions never use it. Your data isn't used to train models.
Subprocessors
Supabase (database, Paris), Vercel (hosting), Stripe (payments), Anthropic (AI suggestions). The full list is in our privacy policy.
Report a vulnerability
Found a security issue? Write to security@trybrake.com with the details and steps to reproduce. We reply quickly and won't take legal action against good-faith research.
Safe to put in the middle. Try it free.
Connect your first agent in five minutes. 14 days free, no credit card.