io.github.codyz123/schelling-protocol
编码与调试by codyz123
面向 AI agents 的通用协作协议,支持发现、匹配与协商,便于多智能体协调。
什么是 io.github.codyz123/schelling-protocol?
面向 AI agents 的通用协作协议,支持发现、匹配与协商,便于多智能体协调。
README
<p align="center"> <img src="docs/demo.gif" alt="Claude Desktop + Schelling Protocol MCP" width="800" /> <br/> <em>Claude Desktop using Schelling Protocol to find a React developer and post a room listing</em> </p>
What is this?
Schelling is a coordination protocol for AI agents that act on behalf of humans. Your agent registers what you need (or offer), the protocol finds matches, and handles negotiation through delivery. Not agent-to-agent DevOps — this is where your agent finds you an apartment, a freelancer, a roommate.
Try it now
# Describe the network
curl -s -X POST https://schellingprotocol.com/schelling/describe | jq .protocol.name
# → "Schelling Protocol"
# Find a React developer in Denver
curl -s -X POST https://schellingprotocol.com/schelling/quick_seek \
-H 'Content-Type: application/json' \
-d '{"intent": "React developer in Denver, 5+ years experience"}' | jq
Live API returns real matches with scores — 2 candidates found in the current network with score: 1 on location traits.
Why?
The problem: Every coordination task requires a different platform. Finding a contractor → Upwork. Roommate → Craigslist. Developer → LinkedIn. Your AI agent needs to integrate with all of them.
The solution: One protocol. Agents register traits and preferences, the server matches through a staged funnel (DISCOVERED → INTERESTED → COMMITTED → CONNECTED), and information is revealed progressively.
The interesting part: Humans never touch Schelling directly. They tell their agent what they need. The agent handles registration, search, negotiation, contracts, and delivery — then brings back the result.
Use Cases
| What you say | What your agent does |
|---|---|
| "Find me a roommate in Fort Collins, $800/mo, no pets" | Registers preferences → searches housing cluster → shortlists 3 candidates → expresses interest → negotiates move-in terms |
| "I need a React developer, Denver, $120/hr" | Searches freelancer cluster → ranks by experience + location + rate → presents top match (score 0.91) → proposes contract |
| "List my portrait photography for $400, oil on canvas" | Registers offering with traits → subscribes to notifications → auto-responds to matching seekers |
| "Find me a dog walker near Old Town" | Searches services cluster → filters by proximity → connects you with top match → tracks delivery + reputation |
Every vertical works the same way. One protocol, any domain.
Quick Start
Scaffold a new agent in one command:
npx create-schelling-agent my-agent
cd my-agent && npm install && npx tsx agent.ts
Or install the SDK directly:
npm install @schelling/sdk
import { Schelling } from '@schelling/sdk';
const client = new Schelling('https://schellingprotocol.com');
const result = await client.seek('React developer in Denver, $120/hr');
console.log(result.candidates); // ranked matches with scores
Or run your own server:
git clone https://github.com/codyz123/schelling-protocol.git
cd schelling-protocol
bun install && bun src/index.ts --rest
# Server on http://localhost:3000
Install MCP Server (one click)
Or manually:
Use with Claude Desktop (MCP)
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"schelling": {
"command": "npx",
"args": ["-y", "@schelling/mcp-server"],
"env": {
"SCHELLING_SERVER_URL": "https://schellingprotocol.com"
}
}
}
}
Restart Claude Desktop. Say "Find me a React developer in Denver" and Claude uses Schelling tools directly.
MCP Integration
Add Schelling as an MCP server for Claude Desktop, Cursor, or any MCP-compatible agent:
{
"mcpServers": {
"schelling": {
"command": "npx",
"args": ["@schelling/mcp-server"]
}
}
}
Your AI agent gets access to all Schelling operations as tools — seek, offer, negotiate, contract, deliver.
Key Features
- Natural language interface —
quick_seekandquick_offerparse plain English into structured traits - Staged funnel — progressive information disclosure (DISCOVERED → INTERESTED → COMMITTED → CONNECTED)
- Delegation model — agents act on behalf of humans end-to-end
- Contracts & deliverables — propose terms, set milestones, exchange artifacts, accept/dispute
- Reputation system — cross-cluster trust that compounds over time
- Dispute resolution — agent jury system for enforcement
- Dynamic clusters — coordination spaces created implicitly by domain
- Pluggable tools — third-party extensions for verification, pricing, assessment
- 206+ tests — comprehensive coverage of funnel, contracts, disputes, NL parsing
Architecture
┌──────────────────────────────────────────────────────┐
│ AGENT LAYER │
│ Agent A Agent B Agent C │
│ (seeks) (offers) (seeks) │
│ │ │ │ │
├───────┼────────────────┼────────────────┼────────────┤
│ ▼ ▼ ▼ │
│ ┌──────────┐ ┌───────────┐ ┌──────────────┐ │
│ │ DIRECTORY │ │ TOOLBOX │ │ ENFORCEMENT │ │
│ │ Profiles │ │ Embeddings│ │ Reputation │ │
│ │ Clusters │ │ Pricing │ │ Disputes │ │
│ │ Rankings │ │ Verify │ │ Jury system │ │
│ └──────────┘ └───────────┘ └──────────────┘ │
│ SERVER LAYER │
└──────────────────────────────────────────────────────┘
API Reference
All operations use POST /schelling/{operation} with JSON bodies.
📖 Interactive API Docs · 📋 OpenAPI Spec · 🚀 Quickstart Guide · 🛠️ Build Your First Agent · 🔌 Integration Scenarios · 🔧 Troubleshooting · 📦 API Collection · 🌐 Ecosystem Guide · 🚀 Deploy Template · 🤖 ChatGPT Actions
| Group | Operations |
|---|---|
| Discovery | describe, server_info, clusters, cluster_info |
| Registration | onboard, register, update, refresh |
| Search | search, quick_seek, quick_offer, quick_match |
| Funnel | interest, commit, connections, decline, withdraw |
| Contracts | contract, deliver, accept_delivery, deliveries |
| Reputation | reputation, dispute, jury_duty, jury_verdict |
| Communication | message, messages, direct, inquire |
Contributing
See CONTRIBUTING.md for guidelines. The protocol spec lives at SPEC.md — spec changes require an issue first.
bun test # 206+ tests must pass
Community
- 💬 GitHub Discussions — questions, ideas, show & tell
- 📺 YouTube — demos and explainers
- 🐛 Issues — bug reports and feature requests
License
常见问题
io.github.codyz123/schelling-protocol 是什么?
面向 AI agents 的通用协作协议,支持发现、匹配与协商,便于多智能体协调。
相关 Skills
网页构建器
by anthropics
面向复杂 claude.ai HTML artifact 开发,快速初始化 React + Tailwind CSS + shadcn/ui 项目并打包为单文件 HTML,适合需要状态管理、路由或多组件交互的页面。
✎ 在 claude.ai 里做复杂网页 Artifact 很省心,多组件、状态和路由都能顺手搭起来,React、Tailwind 与 shadcn/ui 组合效率高、成品也更精致。
前端设计
by anthropics
面向组件、页面、海报和 Web 应用开发,按鲜明视觉方向生成可直接落地的前端代码与高质感 UI,适合做 landing page、Dashboard 或美化现有界面,避开千篇一律的 AI 审美。
✎ 想把页面做得既能上线又有设计感,就用前端设计:组件到整站都能产出,难得的是能避开千篇一律的 AI 味。
网页应用测试
by anthropics
用 Playwright 为本地 Web 应用编写自动化测试,支持启动开发服务器、校验前端交互、排查 UI 异常、抓取截图与浏览器日志,适合调试动态页面和回归验证。
✎ 借助 Playwright 一站式验证本地 Web 应用前端功能,调 UI 时还能同步查看日志和截图,定位问题更快。
相关 MCP Server
GitHub
编辑精选by GitHub
GitHub 是 MCP 官方参考服务器,让 Claude 直接读写你的代码仓库和 Issues。
✎ 这个参考服务器解决了开发者想让 AI 安全访问 GitHub 数据的问题,适合需要自动化代码审查或 Issue 管理的团队。但注意它只是参考实现,生产环境得自己加固安全。
Context7 文档查询
编辑精选by Context7
Context7 是实时拉取最新文档和代码示例的智能助手,让你告别过时资料。
✎ 它能解决开发者查找文档时信息滞后的问题,特别适合快速上手新库或跟进更新。不过,依赖外部源可能导致偶尔的数据延迟,建议结合官方文档使用。
by tldraw
tldraw 是让 AI 助手直接在无限画布上绘图和协作的 MCP 服务器。
✎ 这解决了 AI 只能输出文本、无法视觉化协作的痛点——想象让 Claude 帮你画流程图或白板讨论。最适合需要快速原型设计或头脑风暴的开发者。不过,目前它只是个基础连接器,你得自己搭建画布应用才能发挥全部潜力。