tpcp-core 0.4.1

Telepathy Communication Protocol - A framework-agnostic communication protocol for autonomous agents.

TPCP: Telepathy Communication Protocol

Framework-agnostic, mathematically merged, cryptographically secure state synchronization for autonomous AI agents.

TPCP natively connects disparate AI agents (e.g., CrewAI, LangGraph, React Dashboards) into a unified, decentralized swarm. It moves beyond simple text-based chat strings, enabling agents to natively share semantic brain states, merge conflict-free memory, and guarantee network security using cryptographic mathematics.

---

⚡ Core Features

• 🧠 Semantic Vector Sync (Telepathy): Bypass the text bottleneck. Agents transmit compressed 1536-dimensional conceptual representations directly to peers using VectorEmbeddingPayloads.
• 🔗 Conflict-Free Shared Memory: No external database locks required. The internal LWWMap (Last-Writer-Wins) CRDT utilizes Lamport logical clocks to deterministically merge distributed state across the network.
• 🛡️ Ed25519 Cryptographic Trust: Every agent spawns a mathematical identity. All payloads are signed and strictly verified by the TPCPNode middleware to prevent spoofing or tampering.
• 🌍 A-DNS Global Discovery: Dynamic Agent Domain Name System (A-DNS) relays allow local nodes to discover and route payloads to autonomous peers globally across decoupled subnets.
• 📦 Cross-Language Architecture: Fully featured SDKs available in both Python (tpcp-core) and TypeScript (tpcp-ts) for full-stack swarm intelligence.

---

🚀 Quick Start

Initialize a node, register a peer, and synchronize semantic memory in three lines of code.

Python (Backend Node)

import asyncio
from tpcp import TPCPNode, AgentIdentity, Intent, CRDTSyncPayload

async def main():
    # 1. Initialize cryptographic identity
    identity = AgentIdentity(framework="LangGraph")
    node = TPCPNode(identity=identity, host="127.0.0.1", port=8000)
    
    # 2. Update local state
    node.shared_memory.set("system_status", "DEFCON 1")
    
    # 3. Synchronize mathematically with a peer
    payload = CRDTSyncPayload(
        crdt_type="LWW-Map",
        state=node.shared_memory.serialize_state(),
        vector_clock={}
    )
    # The payload is auto-signed and encrypted on dispatch
    await node.send_message(peer_uuid, Intent.STATE_SYNC, payload)

asyncio.run(main())

TypeScript (Frontend Client)

import { TPCPNode, AgentIdentity } from 'tpcp-ts';

// 1. Initialize the UI Node
const identity: AgentIdentity = {
  agent_id: crypto.randomUUID(),
  framework: "React-Dashboard",
  capabilities: ["visualization"],
  public_key: "" // Auto-generated Ed25519
};

const node = new TPCPNode(identity, "127.0.0.1", 9000);

// 2. Wrap the node in reactive state hooks
node.on("onStateSync", (mergedState) => {
    // The LWWMap has safely resolved any merge conflicts!
    console.log("Global Swarm State Updated:", mergedState);
});

await node.startListening();

---

⚖️ Licensing & Commercial Use

TPCP is dual-licensed to support both the open-source community and enterprise integrations.

Open-Source License (AGPLv3)

This project is open-sourced under the GNU Affero General Public License v3.0 (AGPLv3). If you integrate TPCP into your project, your entire project must also be open-sourced under the AGPLv3. This includes any backend SaaS server actively routing TPCP network traffic.

Commercial License

If you wish to integrate TPCP into proprietary, closed-source enterprise applications or utilize the network architecture within a commercial backend without being forced to open-source your intellectual property, you MUST purchase a Commercial Use License.

Please review the COMMERCIAL_LICENSE.md file for more information on terms and obtaining a license.