Kakunin provides essential KYC (Know Your Agent) compliance infrastructure for AI agents, enabling regulated financial institutions and fintech teams to deploy autonomous AI systems safely and accountably. It addresses the critical need for cryptographic identity, real-time behavioral monitoring, and regulator-ready audit trails for AI agents operating in sensitive environments.
The platform operates through a four-step process:
- Register & Certify: Agents are issued X.509 certificates via AWS KMS, binding their identity, operator, and permitted actions. This process takes less than 3 seconds, establishing a tamper-proof scope for the agent's operations.
- Stream Events: Every agent action, including API calls, decisions, and transactions, is streamed to Kakunin's ingest endpoint at a rate of 1,000 events per second. This creates an immutable audit log automatically.
- Score & Watch: Continuous behavioral drift detection monitors agent activities. Content-risk scoring evaluates what the agent says, not just what it does, flagging harmful or off-scope language. A rolling 30-day trust score is maintained, with automatic revocation if the score drops below a configurable threshold (e.g., 0.85).
- Prove & Report: Kakunin generates regulator-ready compliance reports with a single API call. These reports are mapped to frameworks like NIST AI RMF, MiCA, ISO 27001, and the EU AI Act, providing signed PDF and JSON outputs.
Key features include:
- Cryptographic Boundaries: Utilizes X.509 RSA-2048 certificates and AWS KMS to enforce tamper-proof financial and operational scopes, preventing agents from exceeding limits even if compromised. Counterparties can cryptographically verify agent identities.
- Post-Hoc Proof: Every transaction is signed by the agent, timestamped, and logged immutably, providing a verifiable chain of custody. Behavioral drift detection and auto-revocation ensure immediate response to deviations.
- Content Risk Scoring: Uniquely evaluates agent output for harmful or prohibited content, aligning with EU AI Act Art. 5. High-risk outputs are blocked and logged, offering a crucial layer of protection beyond action logging.
- Real-time Behavioral Kill Switch: Offers continuous rolling 30-day risk scoring and sub-60-second auto-revocation when an agent's behavior deviates from its baseline or crosses a risk threshold. This proactive enforcement prevents breaches and runaway costs.
- Public Verification: Provides a public endpoint where regulators, auditors, or counterparties can verify an agent's identity, scope, operator, and revocation history in under 500ms, without requiring authentication.
- AgentMail: Each certified agent receives a verifiable email address, allowing regulators and counterparties to communicate directly with the agent, with all messages logged in the immutable audit trail.
- API-first Integration: Offers a fully typed TypeScript SDK and OpenAPI 3.0 spec, enabling rapid integration into existing workflows with features like automatic retry, webhook signature verification, and a sandbox mode.
Kakunin is trusted by compliance teams across various sectors, including financial services (e.g., quantitative trading agents, AML/fraud detection), healthcare (diagnostic assistants on EHR), legal (document-review agents), supply chain (automated customs brokers), and the public sector (visa/tax processing agents). It helps organizations pass boardroom validation, scale AI operations, and meet stringent regulatory requirements like the EU AI Act and MiCA, providing a robust solution for AI agent governance.
