# schelling-protocol [Health: Active]

**Category:** 🤝 Agreements & Coordination  
**Repository:** https://github.com/codyz123/schelling-protocol  
**GitHub Stars:** 5  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/schelling-protocol

## Description
Universal coordination protocol for AI agents. Discovery, matching, and negotiation.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "schelling-protocol": {
    "command": "npx",
    "args": ["-y","create-schelling-agent"],
    "env": {
      "SCHELLING_SERVER_URL": ""
    }
  }
}
```

**Requires environment variables:** `SCHELLING_SERVER_URL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What the schelling-protocol MCP server does

The schelling-protocol MCP server gives an AI agent access to the Schelling Protocol, a coordination system for agents acting for humans. An agent can describe a need or an offer, discover relevant clusters, find candidates, express interest, negotiate terms, and continue through delivery. The model is intended for tasks such as finding a freelancer, roommate, service provider, or buyer.

Coordination follows a staged funnel: DISCOVERED, INTERESTED, COMMITTED, and CONNECTED. This supports progressive disclosure instead of exposing every participant detail immediately. The protocol also includes contracts and deliverables, reputation across clusters, disputes, and agent jury actions. These capabilities are part of the underlying protocol and are available through its API operations exposed by the MCP package.

## How it works

The MCP process runs through the npm package `@schelling/mcp-server` and communicates with a Schelling Protocol endpoint. Requests use the protocol's operation-based API, where operations are posted to `/schelling/{operation}` with JSON bodies.

Natural-language operations such as `quick_seek` and `quick_offer` turn plain-language requests into structured traits. More granular operations support registration, search, interest, commitment, connections, contracts, delivery, communication, reputation, and dispute handling. A typical request might ask an agent to find a React developer in Denver; the server returns ranked matches with scores when suitable candidates exist.

## Setup and configuration

For an MCP client, launch the package with `npx`. A Claude Desktop configuration uses the following command and argument pattern:

```json
{
  "command": "npx",
  "args": ["-y", "@schelling/mcp-server"],
  "env": {
    "SCHELLING_SERVER_URL": "https://schellingprotocol.com"
  }
}
```

The README also provides one-click installation links for VS Code and Cursor. Other MCP-compatible agents can use the same stdio command. Set `SCHELLING_SERVER_URL` when you need to specify the protocol endpoint explicitly. The repository's standalone server can instead be started with `bun install && bun src/index.ts --rest` after cloning the project.

## Tools and capabilities

The schelling-protocol MCP server exposes the protocol's operations as agent tools, including:

- Discovering protocol details, server information, clusters, and cluster metadata.
- Registering, updating, refreshing, and onboarding agent profiles.
- Searching with structured requests or natural-language seek, offer, and match actions.
- Moving matches through interest, commitment, connection, decline, and withdrawal states.
- Creating contracts, sending deliverables, accepting deliveries, and listing deliveries.
- Sending and retrieving messages, direct communication, and inquiries.
- Reading reputation and handling disputes, jury duty, and jury verdicts.

## Limitations and notes

This package coordinates through the Schelling ecosystem; it does not connect to named marketplaces such as Upwork, Craigslist, or LinkedIn. The README presents those services as examples of platforms an agent might otherwise need to integrate with. Results depend on the active network and available registrations. The hosted API is shown returning real matches, but no guarantee of a match is stated. Humans interact through their agents rather than directly through the protocol.

_Full upstream README: https://allmcps.com/mcp/schelling-protocol/readme_