# Kakunin

> Cryptographic identity and behavioral accountability for AI agents in regulated industries.

Kakunin issues X.509 certificates to AI agents, monitors their behavior in real-time, and generates compliance reports that satisfy MiCA and EU AI Act audits. Every agent action is logged immutably, signed cryptographically, and auditable by regulators in under 500ms — no account required.

Kakunin is not a human KYC tool (that's Jumio, Onfido, Sumsub). It is the missing compliance primitive for the agents themselves: identity + behavioral proof + auto-revocation at the infrastructure layer.

> Last updated: 2026-05-23

## Product

- [Homepage](https://kakunin.ai): Overview, competitive landscape, regulatory compliance mapping
- [Pricing](https://kakunin.ai/pricing): Starter €195/mo · Pro €1,980/mo · Enterprise custom. 30-day free trial.
- [Documentation Hub](https://kakunin.ai/docs): Full technical and regulatory documentation
- [Core Concepts](https://kakunin.ai/docs/concepts): How identity, behavioral monitoring, and compliance reporting fit together

## Blog

- [Blog](https://kakunin.ai/blog): AI regulation insights, MiCA compliance, EU AI Act guidance
- [circuit-breakers-for-ai-agents](https://kakunin.ai/blog/circuit-breakers-for-ai-agents)
- [why-ai-agents-need-cryptographic-x509-certificates](https://kakunin.ai/blog/why-ai-agents-need-cryptographic-x509-certificates)
- [mica-article-72-ai-agents](https://kakunin.ai/blog/mica-article-72-ai-agents)

## How It Works

- [Agent Registration](https://kakunin.ai/docs/agents): Register an agent via API or SDK — returns agent ID in <1s
- [Certificate Issuance](https://kakunin.ai/docs/certificates): X.509 RSA-2048 cert issued via AWS KMS — private keys never leave the HSM
- [Event Ingestion](https://kakunin.ai/docs/event-ingest): Stream behavioral events at 1,000/s; rolling 30-day risk score updated in real-time
- [Certificate Verification](https://kakunin.ai/docs/verify): Public endpoint — no auth, <500ms p99, returns scope + model hash + revocation history
- [Certificate Enforcement](https://kakunin.ai/docs/enforcement): Auto-revocation when risk score >0.85; fires in <60s; webhook + email alert
- [Webhooks](https://kakunin.ai/docs/webhooks): HMAC-signed event delivery; exponential backoff; zero event loss

## Agent API (Silicon Friendly Level 5 — Autonomous Operation)

Kakunin is rated Silicon Friendly Level 5 — the highest level of AI agent readiness. Autonomous agents can discover, authenticate, and operate against the Kakunin API without any human-in-the-loop setup.

- [Agent Card](https://kakunin.ai/.well-known/agent.json): A2A Agent Card (Google Agent-to-Agent protocol) — 4 published skills
- [MCP Server](https://kakunin.ai/.well-known/mcp.json): Model Context Protocol server card (SEP-1649) — 3 MCP tools
- [Capability Negotiation](https://kakunin.ai/.well-known/agent-negotiate): POST to discover which of 14 API features Kakunin supports for your agent
- [OAuth Discovery](https://kakunin.ai/.well-known/oauth-authorization-server): RFC 8414 — machine-readable auth server metadata
- [OAuth Resource](https://kakunin.ai/.well-known/oauth-protected-resource): RFC 9728 — protected resource metadata
- [SSE Event Stream](https://kakunin.ai/api/v1/agents/{id}/events/stream): GET — real-time Server-Sent Events for agent behavioral events; ?since= cursor support
- [Subscriptions API](https://kakunin.ai/api/v1/subscriptions): POST/GET — programmatic subscription management for 7 event types with lifecycle (create/list/cancel)
- [Notifications Inbox](https://kakunin.ai/api/v1/notifications): GET/POST — proactive inbox for pre-revocation warnings (risk ≥0.75), expiry alerts, quota warnings

## Compliance

- [API Reference](https://kakunin.ai/docs/api-reference): OpenAPI 3.0 spec for all endpoints
- [SDKs](https://kakunin.ai/docs/sdks): TypeScript SDK (@kakunin/sdk) — Zod-validated, sandbox mode, automatic retry. Python V1.1.
- [SLA](https://kakunin.ai/docs/sla): Uptime, latency guarantees, and support tiers
- [CRL](https://kakunin.ai/docs/crl): Certificate Revocation List — public, machine-readable

## Case Studies

- [Case Studies Hub](https://kakunin.ai/docs/case-studies): 6 regulated industries, production deployments
- [Quantitative Trading Agents](https://kakunin.ai/docs/case-study-trading): Tier-1 EU bank · €50M/day scoped in cert · 3× faster execution
- [AML & Fraud Detection](https://kakunin.ai/docs/case-study-aml): Behavioral drift caught misbehavior day 3 · revocation <5ms · $0 fraud loss
- [Diagnostic EHR Assistants](https://kakunin.ai/docs/case-study-healthcare): Read-only cert scope · cannot mutate records · HIPAA trail automatic
- [Document-Review Agents](https://kakunin.ai/docs/case-study-legal): Cryptographic chain of custody for M&A data rooms · court-admissible
- [Automated Customs Brokers](https://kakunin.ai/docs/case-study-supply-chain): KMS-bound signature on declarations · 10× faster clearance
- [Visa & Tax Processing](https://kakunin.ai/docs/case-study-public-sector): EU AI Act-compliant citizen-facing decisions · transparency built in

## Optional

- [AI Discovery Files](https://kakunin.ai/.ai/sitemap.json): Full index of machine-readable content files
- [Complete Product Doc](https://kakunin.ai/llms-full.txt): Full markdown overview for LLMs
- [Brand Guidelines](https://kakunin.ai/.ai/brand-guidelines.txt): Correct terminology, what not to claim
- [FAQ (structured JSON)](https://kakunin.ai/.ai/faq.json): 20 Q&A pairs covering integration, pricing, security, compliance
- [AI Usage Policy](https://kakunin.ai/ai-usage-policy.json): Machine-readable content permissions
- [Agent Skills Index](https://kakunin.ai/.well-known/agent-skills/index.json): Public API skills for agent-to-agent discovery
- [API Catalog](https://kakunin.ai/.well-known/api-catalog): RFC 9727 linkset — service description + docs
- [Agent Card (A2A)](https://kakunin.ai/.well-known/agent.json): Google A2A protocol agent card — 4 skills
- [MCP Server Card](https://kakunin.ai/.well-known/mcp.json): Model Context Protocol server card — 3 MCP tools
- [OpenAPI Spec](https://kakunin.ai/api/v1/openapi.json): Full OpenAPI 3.0 specification — 10+ endpoints