Five-Layer Model
A deep dive into each layer of the HUMΛN architecture
The HUMΛN Protocol is organized into five distinct layers, each building on the one below it. This separation enables modularity, testability, and independent evolution of each component.
Think of it like the OSI model for networking — each layer has a specific job, and higher layers build on lower layers without knowing the implementation details.
Ledger Layer
The foundation of trust. Every action in HUMΛN is logged to an immutable ledger, creating a complete audit trail.
What It Does
- •Records every action with full context (who, what, when, why)
- •Ensures immutability — once written, records cannot be altered
- •Enables provenance queries (e.g., "show me all actions by Agent X")
- •Provides cryptographic proof of action history
Key Concepts
Provenance Record
A single entry in the ledger containing: actor, action, timestamp, resource, delegation chain, and signature.
Distributed Consensus
Multiple nodes agree on the order of records, preventing tampering and ensuring trust.
Identity Layer (Passport)
Establishes cryptographic identity for every entity in the system — humans, organizations, and AI agents.
What It Does
- •Creates unique, verifiable Passports (DID + keys + metadata)
- •Manages delegation tokens for secure, time-limited access
- •Signs actions cryptographically to prove authorship
- •Portable across all HUMΛN apps and services
Key Concepts
Passport
A portable identity containing DID, public/private keys, and metadata (name, email, role, etc.).
Delegation Token
A cryptographic token that allows one Passport to act on behalf of another, with constraints.
Capability Layer (Capability Graph)
Defines what agents can do. Instead of roles ("Admin", "User"), capabilities are fine-grained skills ("invoice_processing", "pdf_extraction").
What It Does
- •Declare capabilities that agents possess
- •Grant capabilities to agents (with expiration and constraints)
- •Verify capabilities before allowing actions
- •Revoke capabilities when no longer needed
Key Concepts
Capability
A named skill or permission (e.g., "invoice_processing", "email_access"). Fine-grained, not coarse.
Capability Grant
A record stating "Passport X has capability Y, granted by Z, until date D".
Orchestration Layer (HumanOS)
Coordinates multi-agent workflows with human-in-the-loop approvals, automatic routing, and intelligent task delegation.
What It Does
- •Break complex tasks into steps and assign to agents
- •Route work based on capabilities, availability, and performance
- •Pause execution for human approval when required
- •Monitor workflow progress and handle failures
Key Concepts
Workflow
A multi-step process involving one or more agents, with defined inputs, outputs, and approval gates.
Human-in-the-Loop
Explicit human approval required before executing high-stakes actions. AI advises, humans decide.
Application Layer
Your code lives here. Build agents, define workflows, integrate with LLMs, and create business logic using all the layers below.
What You Build
- •AI agents with specific skills (invoice processor, support bot, etc.)
- •Custom workflows tailored to your business needs
- •Integrations with your existing systems (CRM, ERP, databases)
- •User interfaces for humans to interact with agents
Start Building
Now that you understand the architecture, explore each layer in detail or jump straight into building.