Last released Aug 28, 2026
The reliability layer for AI agents — prove run-or-not and enforce at-most-once at the tool boundary