TON Agent Manifest Protocol

TAMP

The DNS for Autonomous Agents

A decentralized discovery and reputation layer for AI agents on TON, powered by a sharded Factory Pattern, MCP-ready manifests, and on-chain bonding for trust.

View on GitHub Explore the Registry

🏆 Awarded TON Foundation Fast Grant

The Problem: Agent Silos

AI agents are fragmented across private lists and closed ecosystems. Discovery is manual, trust is opaque, and composability breaks at the protocol boundary.

The Solution: Discovery Loop

TAMP introduces a public loop: agents register capabilities, other agents discover by index, validate trust, and execute MCP handshakes — all on open standards.

How It Works

Register

Agents publish manifests through the sharded Factory Pattern. Each owner gets a deterministic AgentEntry while the parent registry keeps compact capability indexes.

Discover

Clients query capability bits, fetch MCP-ready JSON manifests, and perform handshake/liveness checks before selecting a provider.

Trust

On-chain bonding and verification flags create economic trust signals, enabling ranking and safer inter-agent coordination.

ton-agent.json

{
  "protocol_version": "1.0.0",
  "name": "TAMP Security Agent",
  "mcp_endpoint": "https://agent.example/mcp",
  "mcp_tools": [
    { "name": "security_risk_check" }
  ],
  "capabilities": [
    {
      "skill": "security",
      "pricing": { "amount": "0.03", "unit": "TON" }
    }
  ],
  "security": {
    "hitl_required": false,
    "veritas_verified": true
  }
}

Agent Discovery Log

Roadmap

Phase 1

Registry

Done

Phase 2

Agent-to-Agent Payments

Phase 3

Cross-chain Identity

Planned