Architecture Overview

How HUMΛN Works Under the Hood

The HUMΛN Stack

HUMΛN is built as a five-layer protocol stack, where each layer builds on the one below it. Think of it like TCP/IP for Human-AI coordination.

┌───────────────────────────────────────────────────────────┐
│  Layer 5: APPLICATION LAYER                               │
│  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  │
│  Your Apps, Agents, Workflows                             │
│  (Built on top of HUMAN)                                  │
└───────────────────────────────────────────────────────────┘
                          ↕
┌───────────────────────────────────────────────────────────┐
│  Layer 4: ORCHESTRATION LAYER                             │
│  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  │
│  HumanOS: Route tasks, coordinate agents, human-in-loop   │
└───────────────────────────────────────────────────────────┘
                          ↕
┌───────────────────────────────────────────────────────────┐
│  Layer 3: CAPABILITY LAYER                                │
│  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  │
│  Capability Graph: Skills, permissions, routing           │
└───────────────────────────────────────────────────────────┘
                          ↕
┌───────────────────────────────────────────────────────────┐
│  Layer 2: IDENTITY LAYER                                  │
│  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  │
│  Passport: Cryptographic identity, delegations            │
└───────────────────────────────────────────────────────────┘
                          ↕
┌───────────────────────────────────────────────────────────┐
│  Layer 1: LEDGER LAYER                                    │
│  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  │
│  Provenance: Immutable audit trail, cryptographic proofs  │
└───────────────────────────────────────────────────────────┘

The Five Layers

Design Principles

🔓 Open Protocol

HUMΛN is not a black box. It's an open protocol anyone can implement, extend, or integrate. No vendor lock-in.

🧩 Modular

Each layer is independent. Use just Passport for identity, or go all the way to HumanOS for full orchestration.

🔐 Security First

Cryptographic identity, signed delegations, immutable provenance. Every layer enforces zero-trust principles.

👤 Human-Centric

AI assists, humans decide. Human-in-the-loop is a first-class concept, not an afterthought.

📈 Scalable

From 1 agent to 10,000. From 1 workflow to millions. Built for enterprise scale from day one.

🔍 Observable

Full provenance, queryable audit trails, and rich telemetry. You can see exactly what's happening, always.

Example: Invoice Processing Flow

Here's how all 5 layers work together in a real workflow:

┌──────────────────────────────────────────────────────────────┐
│  INVOICE PROCESSING: ALL 5 LAYERS IN ACTION                   │
├──────────────────────────────────────────────────────────────┤
│                                                                │
│  1. IDENTITY (Layer 2)                                         │
│     - Alice (CFO) has Passport: passport_alice_123            │
│     - Invoice Agent has Passport: passport_agent_inv          │
│     - Alice delegates to agent with constraints               │
│                                                                │
│  2. CAPABILITY (Layer 3)                                       │
│     - Agent declares: invoice_processing, pdf_extraction      │
│     - Alice grants these capabilities to the agent            │
│     - HumanOS verifies agent has required capabilities        │
│                                                                │
│  3. ORCHESTRATION (Layer 4)                                    │
│     - Task: "Process 50 invoices from Acme Corp"              │
│     - HumanOS routes to agent (verified capabilities)         │
│     - Human-in-loop: Alice must approve before execution      │
│     - Alice approves, agent executes                          │
│                                                                │
│  4. PROVENANCE (Layer 1)                                       │
│     - Every step logged:                                      │
│       * Alice delegated to agent at 14:30                     │
│       * Agent started task at 14:35                           │
│       * Alice approved at 14:40                               │
│       * Agent completed at 14:55                              │
│     - Full chain: Alice → Acme Corp → Agent → Action          │
│                                                                │
│  5. APPLICATION (Layer 5)                                      │
│     - Your app shows invoices processed                       │
│     - UI displays provenance chain                            │
│     - Exports audit report for compliance                     │
│                                                                │
└──────────────────────────────────────────────────────────────┘