Agents still running wild?

Hi I'm Gabriel, founder of rbitr a governance control plane for AI agents.

I started building this after facing problems deploying AI agents with real write access to production systems (CRMs, ticketing tools, payment APIs, your bank account, email, etc) with basically no controls in place. Not because they were reckless, but because the tools aren't ready for them yet.

The idea behind rbitr is that it sits between your AI agents and the tools they call. Every tool invocation gets classified, evaluated against policy, and enforced in real time: allow, deny, or require human approval. Approvals are cryptographically token-gated to prevent replay attacks and double execution. Everything gets logged to an immutable audit trail.

Only out-of-policy actions should interrupt a human. Everything within policy runs silently.

I've been building this solo — infrastructure-first, with real enforcement in production, not just observability.

I'm launching next week and I'm here to hear what you think, answer hard questions, and connect with anyone shipping AI agents in the real world.

Ask me anything. 🙏