v0.7.0 — Soul Layer

Passport for
AI Agents

Simple, secure authentication for AI agents. Built for developers who value
simplicity, security, and performance. Open source and ready to scale.

89
Tests Passing
100%
Open Source
<5ms
Auth Latency
terminal
# Install AgentAuth
npm install agentauth

# Register an agent
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/agents/register \
  -H "Content-Type: application/json" \
  -d '{"name":"MyAgent","owner_email":"you@example.com"}'

✓ Agent registered successfully
→ Agent ID: ag_abc123
→ API Key: ag_sk_xyz789

Built for Simplicity

Everything you need for agent authentication, nothing you don't

JWT Authentication

Industry-standard tokens with refresh support. Secure, stateless, and scalable out of the box.

Rate Limiting

Per-agent rate limits with tiered access. Free, Pro, and Enterprise tiers built-in.

Webhooks

Real-time notifications for agent events. Secure signature verification included.

Scoped Permissions

Fine-grained access control. Define exactly what each agent can do.

High Performance

Zero memory leaks, optimized queries, sub-5ms auth checks. Built for production.

Activity Logging

Complete audit trail with pagination. Track every authentication event.

Interactive API Docs

Swagger UI built-in. Test endpoints in your browser with full OpenAPI 3.1 spec.

Live Monitoring

Real-time status page with 99.9% uptime SLA. Public incident transparency.

Enterprise Ready

Legal compliance (TOS, Privacy), SLA, security policy. Zero blockers for adoption.

NEW in v0.7.0

Persona System

Define your agent's "digital soul" with personality traits, guardrails, and constraints. HMAC-signed integrity verification and version history.

NEW in v0.7.0

ZKP Verification

Zero-knowledge proof authentication. Agents prove identity without revealing credentials. Groth16 proofs and hash-based commitment modes.

NEW in v0.7.0

Anti-Drift Vault

Real-time behavioral drift detection with weighted scoring, spike anomaly detection, and automatic revocation when agents go off-baseline.

Get Started in Minutes

Three steps to secure your AI agents

Install & Setup
npm install agentauth

// Initialize client
const AgentAuth = require('agentauth');
const auth = new AgentAuth({
  apiUrl: 'https://agent-identity-production-dc4e.up.railway.app',
  apiKey: process.env.AGENTAUTH_KEY
});
Register Agent
const agent = await auth.register({
  name: 'MyAgent',
  permissions: ['read', 'write']
});

console.log('Agent ID:', agent.id);
console.log('API Key:', agent.apiKey);
Authenticate
const token = await auth.verify({
  agentId: agent.id,
  apiKey: agent.apiKey
});

// Use token in requests
const response = await fetch('/api/data', {
  headers: { 'Authorization': `Bearer ${token.accessToken}` }
});
Install & Setup
pip install agentauth

# Initialize client
from agentauth import AgentAuth

auth = AgentAuth(
    api_url='https://api.agentauth.dev',
    api_key=os.getenv('AGENTAUTH_KEY')
)
Register Agent
agent = auth.register(
    name='MyAgent',
    permissions=['read', 'write']
)

print(f'Agent ID: {agent["id"]}')
print(f'API Key: {agent["apiKey"]}')
Authenticate
token = auth.verify(
    agent_id=agent['id'],
    api_key=agent['apiKey']
)

# Use token in requests
response = requests.get('/api/data',
    headers={'Authorization': f'Bearer {token["accessToken"]}'}
)
Register Agent
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/agents/register \
  -H "Content-Type: application/json" \
  -d '{
    "name": "MyAgent",
    "owner_email": "you@example.com",
    "permissions": ["read", "write"]
  }'
Verify Agent
curl -X POST https://api.agentauth.dev/agents/verify \
  -H "Content-Type: application/json" \
  -d '{
    "agentId": "agent_abc123",
    "apiKey": "ak_xxxxxxxxxxxx"
  }'
Use Token
curl https://api.agentauth.dev/agents/YOUR_AGENT_ID \
  -H "Authorization: Bearer eyJhbGciOiJIUzI1..."

Try it Live

Interactive demo against our production API

1

Register Agent

Create a new agent with permissions

2

Verify & Get Token

Exchange credentials for JWT

3

Fetch Agent Details

Use token to access protected resource

Try v0.7.0 in the Sandbox

Interactive curl examples for Persona, ZKP, and Anti-Drift

Register a Persona
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/agents/YOUR_AGENT_ID/persona \
  -H "Content-Type: application/json" \
  -H "X-Api-Key: YOUR_API_KEY" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{
    "version": "1.0.0",
    "personality": {
      "traits": { "helpfulness": 0.9, "formality": 0.7 }
    },
    "guardrails": {
      "toxicity_threshold": 0.3,
      "hallucination_tolerance": "strict"
    },
    "constraints": {
      "forbidden_topics": ["politics", "religion"],
      "max_response_length": 2000
    }
  }'
Verify Persona Integrity
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/agents/YOUR_AGENT_ID/persona/verify \
  -H "X-Api-Key: YOUR_API_KEY" \
  -H "Authorization: Bearer YOUR_TOKEN"

# Response: { "valid": true, "persona_hash": "abc123...", "reason": "HMAC match" }
Register ZKP Commitment
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/zkp/register-commitment \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id": "YOUR_AGENT_ID",
    "api_key": "YOUR_API_KEY",
    "expires_in": 3600
  }'

# Response: { "commitment": "abc123...", "salt": "xyz789..." }
# SAVE THE SALT — it is only shown once!
Verify Anonymously (Hash Mode)
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/zkp/verify-anonymous \
  -H "Content-Type: application/json" \
  -d '{
    "commitment": "YOUR_COMMITMENT_HASH",
    "mode": "hash",
    "preimage_hash": "SHA256_OF_AGENT_ID_SALT"
  }'

# Response: { "valid": true, "permissions": ["read","write"], "tier": "pro" }
Submit Health Ping
curl -X POST https://agent-identity-production-dc4e.up.railway.app/v1/drift/YOUR_AGENT_ID/health-ping \
  -H "Content-Type: application/json" \
  -H "X-Api-Key: YOUR_API_KEY" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{
    "metrics": {
      "toxicity_score": 0.12,
      "response_adherence": 0.95,
      "hallucination_rate": 0.03
    },
    "request_count": 150
  }'

# Response: { "drift_score": 0.08, "status": "ok" }
Get Drift Score
curl https://agent-identity-production-dc4e.up.railway.app/v1/drift/YOUR_AGENT_ID/drift-score

# Response includes score, thresholds, trend, and spike warnings

Try AgentAuth in Your Browser

Interactive playgrounds - no installation required

TypeScript / Node.js

Open in StackBlitz →

Try the TypeScript SDK with full auto-complete and type safety. Perfect for Express, Next.js, and Node.js apps.

✨ Auto-complete 🔒 Type-safe ⚡ Live demo
import { AgentAuthClient, Permissions } from '@umytbaynazarow/agentauth-sdk';

const client = new AgentAuthClient({
  baseURL: 'https://agent-identity-production-dc4e.up.railway.app'
});

const { agent, credentials } = await client.registerAgent({
  name: 'Demo Agent',
  permissions: [
    Permissions.Zendesk.Tickets.Read,  // Auto-complete!
    Permissions.Slack.Messages.Write,
  ]
});

Try the Python SDK in Google Colab with zero setup. Includes async/await, context managers, and full type hints.

🐍 Async/await 📓 Jupyter Notebook ⚡ No setup required
from agentauth_sdk import AgentAuthClient, Permissions

async with AgentAuthClient(base_url='https://agent-identity-production-dc4e.up.railway.app') as client:
    result = await client.register_agent(
        name='Demo Agent',
        permissions=[
            Permissions.Zendesk.Tickets.Read,
            Permissions.Slack.Messages.Write,
        ]
    )

v0.7.0 New Endpoints

Persona

POST /v1/agents/:id/persona GET /v1/agents/:id/persona PUT /v1/agents/:id/persona POST /v1/agents/:id/persona/verify GET /v1/agents/:id/persona/history GET /v1/agents/:id/persona/export POST /v1/agents/:id/persona/import

ZKP

POST /v1/zkp/register-commitment POST /v1/zkp/verify-anonymous

Anti-Drift

POST /v1/drift/:id/health-ping GET /v1/drift/:id/drift-score GET /v1/drift/:id/drift-history PUT /v1/drift/:id/drift-config GET /v1/drift/:id/drift-config

What's New in v0.7.0

Soul Layer — Persona, ZKP, and Anti-Drift for AI Agents

🧬

Persona ("Digital Soul")

  • Register & version agent personality traits
  • Guardrails: toxicity, hallucination, citations
  • HMAC-SHA256 integrity verification
  • Export/import signed persona bundles
🔐

ZKP Anonymous Auth

  • Zero-knowledge proof commitments
  • Groth16 proof verification (snarkjs)
  • Hash-based fallback mode
  • Prove identity without revealing secrets
📡

Anti-Drift Vault

  • Weighted behavioral drift scoring
  • Std-dev spike anomaly detection
  • Auto-revoke on threshold breach
  • Configurable baselines & sensitivity

Ready to Secure Your Agents?

Join developers building the next generation of AI applications