Cybersecurity professional working on a laptop beside AKRUM encryption code and an activity dashboard

The cryptographic infrastructure layer for the post-quantum era

Encrypt, store and share with quantum-resistant cryptography, and put guardrails on your AI: allow, deny or challenge every action an autonomous AI agent takes.

Start for Free See our AI Agents

Platform

Core technologies. One engine.

Encrypt, Store & Share, and Benchmark, all powered by AKRUM's Cellular Automata entropy engine.

ENCRYPT

Encrypt any file in seconds.

Drop in a PDF, video, spreadsheet, or anything else. AKRUM encrypts it instantly in your browser with quantum-resistant AES-256-GCM, using keys from two independent entropy sources.

Quantum-resistant AES-256-GCM encryption
Files never leave unencrypted
Share with a single link
Tested against NIST STS and Diehard suites

Explore Encrypted File Engine

STORE & SHARE

Post-quantum protected sharing.

Store any sensitive document in an encrypted vault and share it with a single link, protected by our hybrid post-quantum key encapsulation (ML-KEM-768 + X25519).

Post-quantum protected sharing (ML-KEM-768 + X25519)
Quantum-resistant AES-256-GCM encryption
Zero-knowledge, we can’t read your documents
One-click secure share links

Explore Vault + SecureShare

BENCHMARK

Compare entropy sources.

Compare AKRUM’s Cellular Automata entropy engine against OS RNG, Intel RDRAND, and Math.random. See the difference in Shannon entropy, run-length distributions, and byte histograms.

Run AKRUM’s engine against OS RNG, Intel RDRAND, and Math.random
Shannon entropy, run-length, and byte histograms
Export the report for compliance review

Explore Entropy Benchmark Lab

Integrations

Connect AKRUM to your stack.

Auto-encrypt files from Google Drive, Salesforce, and GitHub into your vault, with Slack share alerts and DocuSign document encryption built in.

Google Drive

Available

Auto-encrypt files the moment they land. Connect specific Drive folders and AKRUM encrypts every new file dropped into them. Keys generated by the Cellular Automata entropy engine.

See all integrations

Core API

Post-quantum cryptography, in one REST API.

AKRUM’s REST API gives developers direct access to the same cryptography that powers the platform, Cellular Automata entropy, quantum-resistant encryption, and hybrid post-quantum key encapsulation for protected sharing, from any language, any cloud.

Encrypt, decrypt, generate, and benchmark
Post-quantum key encapsulation (hybrid ML-KEM-768 + X25519)
Cloud-agnostic: runs on AWS, Azure, GCP, or on-prem

Explore Core API

akrum core api

# Cellular Automata entropy
curl -X POST https://akrum.io/api/public/v1/random \
  -H "Authorization: Bearer ak_live_…" \
  -H "Content-Type: application/json" \
  -d '{"bytes":32,"format":"hex"}'

{
  "data": {
    "bytes": 32,
    "format": "hex",
    "value": "b30748143a3ad98e266ff81e…450d4e8",
    "source": "akrum-ca-hkdf-v1"
  },
  "meta": { "algorithm_suite": "envelope-argon2id-aesgcm-v1" }
}

AI Agent Guardrails

Stop your AI agents from going rogue.

AKRUM's AI Agent Guardrails provide safety and policy controls that intercept an agent's execution loop to ensure its actions remain predictable, secure, and compliant.

Autonomous AI agents can access data, use credentials, call tools, and take consequential actions. Most teams can only hope they behave. Our AI Agent Guardrails put an enforceable decision in front of every action.

One enforcement engine.

AI Agent Guardrails is one enforcement engine, a firewall for what your AI agents do. You define guardrails, the rules, and the engine returns allow, deny, or challenge on every action. It isn’t a single fixed rule; it’s one engine running as many guardrails as you need. Anything without a matching rule is denied by default.

Works with any agent.

AI Agent Guardrails works with any autonomous agent, on any framework, because it’s a simple decision API. Wherever your agent is about to do something consequential (access data, use a credential, call a tool, take an action), it calls Guard first and gets back allow, deny, or challenge.

Two kinds of guardrails

Write your own, or turn on ours.

Create custom guardrails from your own policies. Or turn on any of our five prebuilt guardrails, powered by AKRUM's AI Agents. Both kinds run inside the same engine, and a single decision can combine them.

Either way the flow is the same. When your autonomous agent is about to do something, it calls the guardrail first. Inside that one call the engine evaluates your matching guardrails, and for a prebuilt one the evaluation is powered by an AKRUM AI Agent, a deterministic domain engine such as the HIPAA checker or the risk scorer. Those findings become the guardrail's final decision, allow, deny, or challenge, and the decision is signed.

Prebuilt guardrails

Prebuilt guardrails powered by AKRUM's AI Agents.

Five prebuilt guardrails, included in AI Agent Guardrails and powered by our AI Agents. Turn any on, no rules to write.

Healthcare Compliance Agent · Prebuilt guardrail

Continuous HIPAA compliance, over encrypted data.

Automatically check patient-data workloads against HIPAA and HITECH controls, encryption, transmission security, key lifecycle, and audit controls, and get a clear pass/fail verdict with the exact rule and remediation.

HIPAA + HITECH rule evaluation
Specific rule IDs and remediation
Runs without exposing patient data

Explore Healthcare Compliance

{
  "decision": "allow",
  "guardrail": "healthcare-compliance",
  "findings": {
    "verdict": "pass",
    "rules": [
      {
        "id": "HIPAA-164.312-a-2-iv",
        "status": "pass"
      }
    ]
  },
  "summary": "All checks passed. HIPAA, HITECH, GDPR-health."
}

Security Risk Prediction Agent · Prebuilt guardrail

Catch key compromise before it happens.

Score encryption and key events for exfiltration and compromise risk in real time, flagging anomalies and known signatures without ever decrypting the underlying data.

Real-time risk scoring
Anomaly and signature detection
No decryption required

Explore Risk Prediction

{
  "decision": "allow",
  "guardrail": "risk-prediction",
  "findings": {
    "risk_score": 5,
    "matched_signatures": []
  },
  "summary": "Low risk. No anomalous indicators."
}

Encryption Optimization Agent · Prebuilt guardrail

The right cipher for every workload.

Automatically tune cipher, chunk size, and key length for your workload, from real-time streaming to batch backups, balancing throughput, latency, and the security tier you need.

Cipher and chunk-size recommendations
Latency vs. throughput tradeoffs
Tuned per workload and device

Explore Encryption Optimization

{
  "decision": "allow",
  "guardrail": "encryption-optimization",
  "findings": {
    "recommended": {
      "cipher": "AES-256-GCM",
      "chunk_size_bytes": 16384
    }
  },
  "summary": "Optimal cipher selected for this workload."
}

Blockchain Key Manager · Prebuilt guardrail

Automated key lifecycle for on-chain assets.

Evaluate, rotate, and retire wallet, multisig, and validator keys against policy, with HSM-backed rotation plans and clear on-chain transition steps.

Rotation, custody, and recovery
Policy-based lifecycle evaluation
HSM-backed transition plans

Explore Blockchain Key Manager

{
  "decision": "allow",
  "guardrail": "blockchain-key-manager",
  "findings": {
    "rotation_recommended": false,
    "key_type": "hot_wallet"
  },
  "summary": "Key within policy limits. No rotation due."
}

Crypto Compliance Agent · Prebuilt guardrail

Travel Rule, MiCA & FATF screening, built in.

Screen transactions, wallets, and smart contracts against FATF, MiCA, and sanctions lists, with jurisdiction matching and a clear allow / review / block verdict.

FATF Travel Rule + MiCA
Sanctions and smart-contract risk
Jurisdiction-aware verdicts

Explore Crypto Compliance

{
  "decision": "allow",
  "guardrail": "crypto-compliance",
  "findings": {
    "regulations_triggered": [],
    "on_chain_risk_score": 5,
    "sanctions_screening": "clear"
  },
  "summary": "Clear. No FATF, MiCA, or sanctions hits."
}

Deterministic & signed

AI assists. The engine decides. Every decision is sealed.

An AI model helps interpret a request and extract parameters, but it never makes the security call. A deterministic engine owns every allow, deny, or challenge. Each decision is signed with post-quantum ML-DSA (FIPS 204) and written to a tamper-evident log anyone can verify offline with the public key.