AgentPass | AI Agent Identity and Trust Platform

AgentPass

AI Agent Identity and Trust Platform

You control the agent. The agent does not control you.

ECDSA P-256MCPS SIGNEDTRUST L0–L4 OFAC SDNSPIFFEHASH-CHAINED LEDGER

Password

Forgot password?

AgentPass -- AI Agent Identity and Trust Platform

CyberSecAI Ltd. Registered in England and Wales.

CyberSecAI^® is a registered trademark of CyberSecAI Ltd.

IANA PEN 66339 | Patent Pending (PCT) | Four UK Patents Filed

contact@cybersecai.co.uk

Deployment Successful

Tenant ID

API Key

Endpoint

Agents Active

Actions Signed

Actions Blocked

Certs Issued

Sanctions Screened

Ledger Entries

Recent Activity

Agent Health

Agents

Manage agent identities, certificates, and trust levels.

Name	SPIFFE ID	Trust	Status	Capabilities	Created	Actions

Trust Centre

Graduated trust levels from L0 (anonymous) to L4 (critical with human approval).

L0

Anonymous / Unverified

read_data
No signing
Rate limited

L1

Verified Identity

query_database
call_api
Cert issued
Actions signed

L2

Trusted Operator

write_file
send_message
MCPS signing
Sanctions screened

L3

Payment Authorised

pay
transfer
Human approval opt-in
Reason required

L4

Critical / Human-in-Loop

deploy
Mandatory human approval
AgentBee hardware gate
Full audit trail

Policy Rules

Action	Required Level	Approval
read_data	L0	Auto
query_database	L1	Auto
call_api	L1	Auto
write_file	L2	Auto
send_message	L2	Auto
pay	L3	Human opt-in
transfer	L3	Human opt-in
deploy	L4	Human required

Promotion Log

Timestamp	Agent	From	To	By	Reason

Evidence Ledger

Tamper-evident, hash-chained audit trail for every agent action.

#	Timestamp	Type	Agent	Action	MCPS Nonce	Result	Hash

Compliance

Framework mapping and compliance status for regulated deployments.

EU AI Act COMPLIANT

Art 12 -- Record-keeping (evidence ledger)
Art 14 -- Human oversight (L4 gate)
Art 15 -- Accuracy, robustness, cybersecurity

SOC 2 COMPLIANT

Security -- mTLS, SPIFFE, cert-based auth
Availability -- multi-region PoP
Processing Integrity -- MCPS signing

Korean AI Basic Act COMPLIANT

Agent identification requirements
Risk assessment documentation
Audit trail preservation

GDPR COMPLIANT

Art 5 -- Data minimisation
Art 25 -- Privacy by design
Art 30 -- Processing records

Control Mapping

Requirement	AgentPass Capability	Status	Evidence
EU AI Act Art 12	Evidence Ledger (hash-chained)	PASS	Ledger export
EU AI Act Art 14	L4 Human-in-Loop gate	PASS	Kill Centre log
EU AI Act Art 15	MCPS signing + AEBA monitoring	PASS	Signed action receipts
SOC 2 Security	mTLS, SPIFFE, KMS encryption	PASS	CA cert chain
SOC 2 Availability	Multi-region PoP deployment	PASS	Region status
SOC 2 Processing	MCPS per-action signing	PASS	Nonce + signature
Korean AI Basic Act	Agent identity + audit trail	PASS	SPIFFE certs + ledger
GDPR Art 25	Privacy-by-design architecture	PASS	Architecture doc

Monitoring

Real-time agent activity feed, alerts, and behavioural analytics.

High Frequency

payment-processor-01 executed 47 actions in last 5 minutes

Repeated Blocks

data-scraper-03 blocked 12 times (insufficient trust level)

Live Event Feed

Agent Activity Heatmap

Sanctions Screening

Total screened0

Matches found0

Last check--

Kill Centre

Instantly revoke agent certificates and permissions. Irreversible.

Revocation Log

Timestamp	Agent	Revoked By	Reason

Emergency Kill All

Immediately revoke ALL agent certificates. This cannot be undone.

CA Settings

Certificate Authority configuration and status.

Certificate Authority

AgentPass CA

OISTE Root of Trust

Status

CONNECTED

Region

Configured

Certificate Profile

AIC (OID 1.3.6.1.4.1.66339.1)

Total Issued

Active

Revoked

SDK & API

Integration guides, connection details, and API reference.

Endpoint

API Key **********************

Region

CA Certificate Authority

Code Snippets

Python (agentpass-sdk)

from agentpass import AgentPass

ap = AgentPass(
    endpoint="",
    api_key="YOUR_API_KEY",
    region=""
)

agent = ap.create_agent(
    name="payment-processor",
    trust_level=2,
    capabilities=["pay", "call_api"]
)

receipt = agent.execute("pay", {
    "amount": 100,
    "currency": "USD",
    "counterparty": "vendor-01"
})
print(receipt.mcps_nonce)

Node.js (@agentpass/sdk)

import { AgentPass } from '@agentpass/sdk';

const ap = new AgentPass({
  endpoint: '',
  apiKey: 'YOUR_API_KEY',
  region: ''
});

const agent = await ap.createAgent({
  name: 'data-analyst',
  trustLevel: 1,
  capabilities: ['read_data', 'query_database']
});

const result = await agent.execute('query_database', {
  query: 'SELECT * FROM transactions LIMIT 10'
});
console.log(result.mcpsNonce);

Claude Code MCP Config

{
  "mcpServers": {
    "agentpass": {
      "command": "npx",
      "args": ["-y", "@agentpass/mcp-server"],
      "env": {
        "AGENTPASS_ENDPOINT": "",
        "AGENTPASS_API_KEY": "YOUR_API_KEY",
        "AGENTPASS_REGION": ""
      }
    }
  }
}

curl

curl -X POST /api/agents \
  -H "X-Api-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-agent",
    "trustLevel": 1,
    "capabilities": ["read_data", "call_api"]
  }'

API Reference

POST/api/registerRegister new tenant
POST/api/loginAuthenticate and get API key
GET/api/agentsList all agents
POST/api/agentsCreate a new agent
GET/api/agents/:idGet agent details
POST/api/agents/:id/executeExecute a signed action
PUT/api/agents/:id/promotePromote agent trust level
PUT/api/agents/:id/demoteDemote agent trust level
DELETE/api/agents/:idKill (revoke) an agent
GET/api/ledgerGet evidence ledger entries
POST/api/ledger/verifyVerify ledger integrity
GET/api/statsGet tenant statistics
GET/api/ca/statusCA connection status
POST/api/sanctions/screenScreen a name against sanctions
GET/api/agents/:id/statusPublic agent status page

Security & MFA

Multi-factor authentication, password management, and security settings.

Multi-Factor Authentication

TOTP Authenticator App

Compatible with Microsoft Authenticator, Google Authenticator, Authy

AgentPass Mobile Authenticator BETA

Hardware-backed ECDSA signing from your phone. Required for L4 approvals.

Hardware Security Key (FIDO2)

YubiKey or compatible hardware key

Coming Soon

Password Management

Current Password

New Password

Confirm New Password

Minimum 12 characters. Must include uppercase, lowercase, number, and special character.

Protocol Binding

Select which signing protocol agents use. This applies to all new actions.

MCP (MCPS) -- Signs JSON-RPC messages via mcp-secure. Use when agents communicate over the Model Context Protocol.

REST (ATTP) -- Signs HTTP requests via X-ATTP-* headers (draft-sharif-attp-01). Use when agents call REST APIs directly.

Both -- Dual mode for migration or hybrid environments. Returns both MCPS envelope and ATTP REST headers.

Active Sessions

Current Session

Browser -- ${navigator.userAgent.includes('Chrome') ? 'Chrome' : navigator.userAgent.includes('Safari') ? 'Safari' : 'Browser'}

Active

Billing

Plan details, usage, and upgrade options.

Usage This Period

Agents

Actions

Certificates

Plans

Starter

$600/mo

10 agents, MCPS signing, evidence ledger, CA certificates, email support

Enterprise

$2,500/mo

Unlimited agents, geo-lock jurisdictional binding, mobile authenticator, custom CA, SLA, priority support

Unlimited

Custom

On-premise deployment, air-gapped, HSM integration, bespoke compliance, dedicated account manager

Licence Key Required. All plans require a valid licence key issued by CyberSecAI Ltd. The licence key is cryptographically signed and cannot be generated without authorisation.