create-atp-agent

Bootstrap a quantum-safe AI agent in under 60 seconds with an embedded onboarding dashboard.

Zero Install (Node.js auto-detected)

Don't have Node.js? No problem — the installer handles everything:

Mac / Linux:

curl -fsSL https://agenttrustprotocol.com/install.sh | bash

Windows (PowerShell as Admin):

irm https://agenttrustprotocol.com/install.ps1 | iex

Already have Node.js?

npx create-atp-agent

That's it. This will:

1. Ask for your agent name
2. Scaffold a ready-to-run project (agent.ts, package.json, README.md)
3. Launch an embedded onboarding dashboard in your browser at http://localhost:3456
4. Generate a quantum-safe DID locally (no backend services required)

Embedded Onboarding Dashboard

After scaffolding, a local web server starts automatically and opens a step-by-step onboarding wizard in your browser:

• Step 1: Runtime - Select your agent runtime (OpenClaw, MCP, LangChain, ADK, or custom)
• Step 2: Agent Name - Pre-filled from your CLI input
• Step 3: Security Profile - Choose from 4 built-in profiles (safe-default, dev-mode, enterprise-locked, openclaw-sandbox)
• Step 4: Confirm - Review your configuration and complete registration

The dashboard uses a dark-mode UI matching the ATP web application design. No external dependencies needed - runs entirely on Node.js built-ins.

What Gets Created

my-agent/
  |-- agent.ts         <- your agent (runs offline!)
  |-- package.json
  |-- README.md

Standalone Mode (Offline-First)

The generated agent works immediately without any backend services:

import { Agent } from 'atp-sdk';

// One line — creates agent and prints status automatically
await Agent.quickstart('MyAgent');
// ⚡ MyAgent ready!
//   DID:          did:atp:a1b2c3...
//   Quantum-safe: yes
//   Mode:         standalone (local)

When ATP services aren't running, Agent.create() automatically falls back to standalone mode:

• DID is generated locally from the cryptographic keypair
• All quantum-safe crypto works offline
• No 30-second timeout, no crash, no errors

When you're ready for full features (trust scoring, verifiable credentials, identity registration), start the ATP services with docker-compose up -d.

Security Profiles

Each profile defines what your agent can and cannot do:

| Profile | Shell | Filesystem | Network | Best For | | --- | --- | --- | --- | --- | | safe-default | Blocked | Read-only | Internal only | Most agents | | dev-mode | Allowed | Read + Write | All domains | Local development | | enterprise-locked | Blocked | Approved paths | Internal corp | Production | | openclaw-sandbox | Allowlist only | Sandbox paths | Internal + partners | OpenClaw agents |

CLI Commands

# Scaffold a new agent (interactive)
npx create-atp-agent

# Alternative alias
npx atp-init

# Onboard an existing agent via CLI wizard
npx atp-onboard-agent

Requirements

• Node.js 18+
• npm 7+

Next Steps After Setup

cd my-agent
npm install
npm start

Then:

• Visit the onboarding dashboard at http://localhost:3456 to configure your security profile
• Run docker-compose up -d from the atp-core repo for full network features
• Read the ATP SDK docs

License

Apache-2.0

Agent Trust Protocol - The security layer for AI agents. Built by Sovr INC.