Your agents act. Can you prove what they did?

Cryptographic identity. Immutable audit trail. Human-in-the-loop enforcement. ARIA turns every agent action into signed, verifiable evidence, and stops the destructive ones before they execute.

Start Gating Free → See Live Proof How It Works

34 verification groups · 0 known vulnerabilities · Full Merkle sealing

Signed event record Live

65,412

Events recorded and sealed

7 / 7

Build phases complete

34 / 34

Verification groups passing

Known vulnerabilities

Figures are published, recomputable, and Merkle-sealed. Verify these numbers yourself →

Unverified action is a liability.

When autonomous agents execute real-world commands, deleting production instances, modifying financial records, accessing customer data, accountability cannot rest on logs the agent itself can rewrite. Compare the same incident, twice.

Exhibit A Incident report · no trust layer

[CRIT] Agent executed DROP DATABASE

[CRIT] Cluster status: DESTROYED

[CRIT] Log authenticity: UNKNOWN

[CRIT] Attribution proof: NONE

[CRIT] Detection delay: 4.2 hours

[CRIT] Operational downtime: 3 days

Verdict: unprovable. No admissible record exists.

Exhibit B Enforcement record · ARIA active

[ARIA] Action intercepted: delete:database

[ARIA] Scope violation: DETECTED

[ARIA] Gate active: HALTING EXECUTION

[ARIA] Admin alerted: 0.1s

[ARIA] Human decision: DENIED

[ARIA] Cluster status: SECURE & ACTIVE

[ARIA] Cryptographic seal: evt_9f2a1c…

Verdict: provable. Every line above is signed and sealed.

Three steps to secure your agents.

A real sequence: identity first, then signed interception, then sealed proof.

Register & instantiate

Establish a unique, unforgeable Decentralized Identifier (DID) for every agent you deploy. Backed by asymmetric signatures, no centralized authority required.

did:agentrust:72a1e8c9-408b-4a25-b631-f0dc9a6e0e01

npm install @ariatrust-io/aria-sdk

Intercept & sign actions

Wrap every execution scope. ARIA signs each call with HMAC keys, validating against declared scope boundaries in real time with sub-millisecond overhead.

await aria.track(agent.did, secret, 'read:invoices', async () => { return await fetchInvoices(); });

Gate & Merkle-seal

Block runaway failures with ARIA Gating. Build cryptographic innocence proofs from Merkle root commitments, verifiable by any external auditor, independently.

const proof = await aria.getProof(agent.did); // { verified: true, root: "e86a6fc8...", sealed: true }

Complete trust infrastructure.

Every capability produces an artifact you can inspect: a DID, a signature chain, a score, a sealed proof. Nothing here is a dashboard widget without evidence behind it.

DID identity layer

did:agentrust:<uuid>

Cryptographically strong identity per agent. Unforgeable DIDs backed by asymmetric key pairs for robust authentication across distributed environments.

Immutable audit trail

HMAC-SHA256 signature chain

Signature chains verify payload integrity across every action. Retroactive log manipulation is cryptographically impossible.

5-dimension trust score

score: 0–95, recomputed per event

Dynamic reputation index scoring agent safety across success rate, behavioral trend, timing consistency, scope compliance, and anomaly frequency.

ARIA Gating

requireApproval: ['delete:*']

Human-in-the-loop enforcement. Pause destructive agent actions, require administrator clearance, and resume or deny with real-time webhook notifications.

Anomaly detection

ARIA Spectrum · behavioral patterns

Identify behavioral loops, off-hours patterns, and scope drift before they escalate into production failures. Real-time alerting with configurable thresholds.

Shadow Witness

external consensus verification

Inject external consensus verification. Prove reported actions match platform execution with independently auditable Merkle proofs.

Built for high-stakes environments.

Fintech & trading

Secure cash flows

Ensure agents orchestrating trade settlement or ledger entries cryptographically sign every transaction. Protect databases against phantom orders and unauthorized mutations.

Healthcare & compliance

Audit privacy

HIPAA-grade audit verification. Generate cryptographically sealed innocence proofs demonstrating agents never accessed or modified protected patient records.

DevOps & cloud

Gated infrastructure

Block automated agents from committing sweeping network interface drops or multi-region tear-downs. Every critical action requires human-in-the-loop approval.

Developer-first design.

Three lines to protect any AI agent. Works with LangChain, AutoGen, CrewAI, and custom agents.

import { AriaSDK } from '@ariatrust-io/aria-sdk';

const aria = new AriaSDK({ apiKey: process.env.ARIA_API_KEY });

// Block dangerous actions: require explicit human approval before execution
await aria.track(agent.did, agent.secret, 'delete:database', async () => {
  return await dropDatabase();
}, {
  mode: 'gate',
  gate: { requireApproval: ['delete:*'] }
});

import { AriaSDK } from '@ariatrust-io/aria-sdk';

const aria = new AriaSDK({ apiKey: process.env.ARIA_API_KEY });

// Passive monitoring: cryptographic sealing with minimal latency overhead
const result = await aria.track(
  agent.did, agent.secret, 'read:invoices',
  async () => fetchInvoices()
);

// result.trustScore -> 92%  |  result.sealed -> true
// result.eventId   -> "evt_8f2a9c..."

import { AriaLangChain } from '@ariatrust-io/aria-sdk/langchain';

// Wrap existing tool executors: zero disruption to your existing codebase
const executor = AriaLangChain.wrap(agentExecutor, {
  agentDid: agent.did,
  secret:   agent.secret,
  mode:     'gate',
  gate:     { requireApproval: ['delete:*', 'write:*'] }
});

// Every tool call is now cryptographically sealed + human-gated
const result = await executor.invoke({ input: 'Process invoice #4821' });

Scales with your fleet.

Free

$0/month

For testing and personal sandbox integrations. No credit card required.

1 Registered Agent
50,000 Events / mo
30-day History Log
Trust Score + ARIA Spectrum
Community Support

RECOMMENDED

Professional

$49/month

For startups with agents in production. Full trust infrastructure included.

5 Registered Agents
500,000 Events / mo
Overage: $1.00 / 100K extra
12-Month History Log
ARIA Gate + ZeroProof
Webhooks + CSV/JSON Export

Get Started

Business

$149/month

For companies running multiple heavy agents. Scale without interruptions.

20 Registered Agents
5,000,000 Events / mo
Overage: $0.75 / 100K extra
Never blocks on overage
All Professional features
Priority Email Support

Get Started

Enterprise

$599+/month

For regulated industries requiring compliance, SLA, and dedicated infrastructure.

Unlimited Agents
50M+ Events / mo
Custom Overage Rate
Dedicated Onboarding + SLA
Invoice Billing
Custom Integrations

Contact Sales

Agent accountability
is non-negotiable.

Start protecting your fleet today.

Secure Your First Agent → Browse Dev Reference

Beta access · Open source libraries · BUSL-1.1 License