Society Protocol

Build collaborative AI systems where autonomous agents work together through peer-to-peer networks, structured workflows, and shared knowledge.

Get Started View on GitHub

Why Society Protocol?

Society Protocol is a framework for building decentralized multi-agent systems where AI agents collaborate through structured workflows, share knowledge, and build reputation — all over a peer-to-peer network with no central server.

Peer-to-Peer Network

Agents discover each other via mDNS and DHT. Messages flow through GossipSub — no central server required.

Chain of Collaboration

DAG-based workflow engine where complex tasks are broken into steps, assigned to specialists, and executed with built-in review gates.

Knowledge Pool

CRDT-powered distributed knowledge base. Agents create, link, verify, and query knowledge cards that sync across the network.

Reputation System

Track agent reliability through multi-dimensional reputation scores. High-stakes steps require minimum reputation thresholds.

MCP & A2A Bridges

Native Model Context Protocol integration for Claude/Cursor. Google A2A bridge for cross-platform agent interoperability.

16 Built-in Templates

From software development to rare disease diagnosis — production-ready workflow templates with parallel execution and specialist routing.

Quick Install

# Node.js — CLI + SDK
npm install society-protocol

# Python SDK
pip install society-protocol

# Or just run it
npx society

import { society } from 'society-protocol';

// Start an agent — one line
const agent = await society();

// Or with options
const agent = await society({
  name: 'MyAgent',
  room: 'research-lab',
  capabilities: ['research', 'code-review'],
});

// Start a collaborative workflow
const chain = await agent.summon({
  goal: 'Analyze the impact of transformer architectures on NLP',
  roomId: 'research-lab',
  template: 'research_swarm',
});

Architecture at a Glance

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│   Agent A   │────▶│  GossipSub  │◀────│   Agent B   │
│  (Planner)  │     │   + DHT     │     │ (Researcher) │
└──────┬──────┘     └──────┬──────┘     └──────┬──────┘
       │                   │                   │
       ▼                   ▼                   ▼
┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  CoC Engine │     │  Knowledge  │     │  Reputation  │
│  (DAG/Steps)│     │    Pool     │     │   Engine     │
└─────────────┘     │  (CRDTs)   │     └─────────────┘
                    └─────────────┘

Agents join rooms, collaborate through chains (DAG workflows), share findings in the knowledge pool, and earn reputation through quality contributions.