Current Work KYA protocol architecture Agent audit logs (on-chain + off-chain) Behavioral monitoring systems Writing AI agents in finance Accountability > trust Why MCP is not enough