Open source · evidence over claims

Trusted infrastructure for autonomous AI.

We build tooling that makes AI-written code prove itself: an executed check instead of a sentence, and a ledger anyone can recompute.

Projects

CUSTOS

Proof-based guardrails for AI coding agents: no green check, no “done”. Claude Code plugin, MIT.

Selfproof

Checks AI-written code and records evidence on a tamper-evident ledger. Pre-release, Apache-2.0.

Stance

Honest by design. Prototypes are called prototypes. We never claim “unhackable” or “bug-free”; the strongest claim is zero known findings per named tools on a date.