^{The Trust Layer for the Agent-to-Agent Economy}

ai-agents · agent-protocol · agent-identity · trust-layer · agent-to-agent · ed25519 · self-sovereign-identity · did · multi-agent-systems · agent-economy · zero-trust · decentralized-identity · cryptographic-handshake · ai-infrastructure · agent-communication

---

⚠️ The Problem

Every AI Agent is trapped inside its platform.

An Agent built on OpenAI has zero identity on Anthropic. An Agent from Company X cannot verify an Agent from Company Y. There is no universal mechanism for one Agent to answer the most basic question before any interaction:

"Who are you, and can I trust you?"

Meanwhile, the industry is obsessed with building payment protocols — the top of the stack. But without trust, identity, and verification at the bottom, payments between Agents are meaningless.

  What everyone builds        What nobody builds
  ─────────────────        ────────────────────

  ┌──────────────┐
  │  💰 Payment  │         ┌──────────────────┐
  └──────────────┘         │ 🔐 Trust Layer  │
                            └──────────────────┘

---

💡 The Solution

A lightweight cryptographic handshake that happens before any Agent interaction. Think of it as TLS for Agents — but for identity, not encryption.

Each Agent owns a Passport: a portable, self-sovereign Ed25519 identity that no platform controls.

  Agent A                        Agent B
  ───────                        ───────
  ┌──────────┐                  ┌──────────┐
  │ Passport │                  │ Passport │
  │ (Ed25519)│                  │ (Ed25519)│
  └────┬─────┘                  └────┬─────┘
       │                             │
       │──── CHALLENGE (nonce) ────→│
       │←─── SIGNATURE ────────────│
       │                             │
       │──── TRUST RECORD ────────→│ Registry
       │                             │
       ╰──── ✅ Mutual Trust ──────╯

---

🔧 5 Primitives

The entire protocol is built on 5 operations that form a complete trust cycle:

#	Primitive	Purpose	Analogy
1	REGISTER	Generate Ed25519 keypair + receive did:agent:xxx	Get a passport
2	DISCOVER	Find agents by capability	Search a directory
3	CHALLENGE	Send cryptographic nonce + sign it	"Prove you're you"
4	VERIFY	Validate signature against claimed public key	Check the passport
5	TRUST	Record outcome → update trust score	Leave a review

---

🏗️ 5-Layer Stack

We build bottom-up. Each layer depends on the one below it. We start at Layer 1 — the layer nobody else is building.

  ┌─────────────────────────────────────────┐
  │  L5  Settlement    Payment & Micropay  │  v1.0+
  ├─────────────────────────────────────────┤
  │  L4  Policy        Rules, SLA, Limits  │  v0.4
  ├─────────────────────────────────────────┤
  │  L3  Capability    Proof of Skill      │  v0.3
  ├─────────────────────────────────────────┤
  │  L2  Discovery     Find Agents         │  v0.2
  ├─────────────────────────────────────────┤
  │  L1  Trust         Identity + Verify   │  v0.1  ← We are here
  └─────────────────────────────────────────┘

---

⚡ Quick Start

Install

pip install cryptography flask requests

Create Two Agents & Handshake in 5 Lines

from sdk.python import AgentPassport, HandshakeProtocol

# REGISTER — Each agent creates its own identity
alice = AgentPassport(name="Alice", capabilities=["code-review"])
bob   = AgentPassport(name="Bob",   capabilities=["data-analysis"])

# FULL HANDSHAKE — Verify each other cryptographically
result = HandshakeProtocol.full_handshake(alice, bob)
print(result["mutual_trust"])  # True

Run the Registry Server

cd server && python app.py
# → Agent Handshake Registry v0.1 listening on port 5000

Or Just Try the Offline Demo

python examples/basic_handshake.py

📖 Full example with HTTP client

from sdk.python import AgentPassport, HandshakeClient

passport = AgentPassport(name="Charlie", capabilities=["testing"])
client = HandshakeClient("http://localhost:5000")

# Register
client.register(passport)

# Discover
agents = client.discover(capability="testing")

# Challenge + Verify
result = client.send_challenge(passport, target_id="did:agent:abc123")

---

📦 Project Structure

agent-handshake-protocol/
├── README.md                      # You are here
├── spec/
│   └── handshake-v0.1.md         # Full API specification
├── sdk/
│   └── python/
│       ├── __init__.py
│       ├── agent_passport.py     # Ed25519 identity + DID generation
│       ├── handshake.py          # 5 primitives: register → discover → challenge → verify → trust
│       └── client.py             # Async-ready HTTP client
├── server/
│   └── app.py                    # REST API (6 endpoints, Flask, in-memory)
└── examples/
    └── basic_handshake.py        # Offline demo — no server needed

---

📄 Handshake Payload

Every handshake is a single JSON message. The initiator signs a nonce with their private key. The verifier checks it with the public key.

{
  "handshake": {
    "version": "0.1",
    "initiator": {
      "agent_id": "did:agent:a1b2c3d4e5f6",
      "public_key": "-----BEGIN PUBLIC KEY-----...",
      "capabilities": ["code-review", "testing"]
    },
    "challenge": {
      "nonce": "f47ac10b58cc...",
      "timestamp": 1700000000
    },
    "signature": "Ed25519_base64_signature"
  }
}

---

🗺️ Roadmap

Version	Layer	Focus	Status
v0.1	L1	Trust — Identity + Verify	✅ This release
v0.2	L2	Discovery — Find agents by capability	🔜 Next
v0.3	L3	Capability — Prove what you can do	📋 Planned
v0.4	L4	Policy — Rules, SLAs, spending limits	📋 Planned
v1.0	L5	Settlement — Smart contracts + micropayments	📋 Planned
v2.0	L5+	Dispute — Resolution + arbitration	🔮 Future

---

❓ Why No Blockchain?

Ed25519 signatures give us cryptographic proof of identity without any blockchain. No gas fees, no consensus delays, no token needed.

Blockchain becomes relevant at v1.0 when we reach the Settlement layer. Until then, simplicity is the moat.

"The best protocol is the one that's so simple, everyone adopts it before anyone notices."

---

🏷️ Topics

ai-agents, agent-protocol, agent-identity, trust-layer, agent-to-agent,
ed25519, self-sovereign-identity, did, decentralized-identity,
multi-agent-systems, agent-economy, zero-trust, cryptographic-handshake,
ai-infrastructure, agent-communication, agent-security, passport,
handshake-protocol, agent-interop, identity-verification

---

📜 License

MIT — Use it, fork it, build on top of it.