With AI agents getting more autonomous, controlling irreversible actions (sending money, emails, deploying code) is becoming critical.

Nyxi introduces execution-time governance: a clean veto/allow boundary that works regardless of whether proposals come from humans or models.

Public docs and demos here (proprietary, no source): https://github.com/indyh91/Nyxi-Showcase

Main overview: https://github.com/indyh91/Nyxi-Showcase/blob/main/docs/PROD...

Would love feedback on the concept!