Last released Jun 23, 2026
Grounding firewall for action-taking AI agents — blocks hallucinated tool arguments and fabricated spoken facts in code, before they cause harm. Zero dependencies, model-agnostic.
Supported by
