raucle

The graph and the audit export for your AI agents.

Built on the open-source raucle-detect engine, with formal proofs you can hand to a regulator.

What it does

Drop in your agent. See every tool call it could be tricked into making. Get a PDF audit report — green nodes are mathematically safe, red nodes are open findings — that you hand to your FCA, BaFin, or EU AI Act examiner.

Built for CISOs, Heads of AppSec, and GRC leads at fintechs and AI-native startups who need to prove their agents are safe before they touch real customer data.

Why this exists

AI agent security today is vendor heuristics and prompt-engineering folklore. None of it would survive the kind of scrutiny a payments system or a government workload gets. raucle is built to the standard regulated industries already demand for everything else — formal, mechanised, provable.

Research

Verified Capability Discipline for LLM Agent Tool Calls. Under submission to IEEE S&P 2027. Branded preprint coming soon — request a copy.

Status

Live raucle-detect — the open-source engine. Active development at github.com/craigamcw/raucle-detect.
Soon raucle Cloud — the graph and the audit export. Private development. Join the waitlist for early access.

Join the waitlist

We'll only email you when there's something real to show. No marketing list, no drip campaign.