mcpField
M
mcp-server-fear-greed

mcp-server-fear-greed

Local
github.com/ycjcl868Finance4Updated 59w ago

About mcp-server-fear-greed

mcp-server-fear-greed provides the CNN Fear & Greed Index for US markets to financial workflows and MCP clients. It returns structured and text content, including current readings, historical comparisons, and component indicators.

Key features

  1. 01Fetch the current sentiment index
  2. 02Compare prior close, week, month, and year
  3. 03Inspect individual market indicator scores
  4. 04Return structured Markdown or JSON

Use cases

  1. 01Check market sentiment during financial analysis
  2. 02Compare current readings with historical periods
  3. 03Pass market indicators to an MCP client

MCP support

The MCP server exposes the CNN Fear & Greed Index and related market indicators through the `get_fear_greed_index` tool. It supports stdio, in-process calls, and remote SSE or Streamable HTTP connections.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"ycjcl868-mcp-server-fear-greed": {
"command": "npx",
"args": [
"-y",
"mcp-server-fear-greed"
]
}
}
}
Run commandnpx -y mcp-server-fear-greed

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmmcp-server-fear-greedlatest

Tools

1

Saved tool list · updated 2w ago

This tool takes no parameters.

Details

From the project README

MCP Server Fear & Greed Index

NPM Downloads

Install MCP Server

A Model Context Protocol (MCP) server that provides access to the CNN Fear & Greed Index for the US stock market. This server fetches real-time market sentiment data and presents it in both structuredContent and text content.

README image

Features

  • Real-time Fear & Greed Index: Get the current market sentiment score (0-100)
  • Historical Comparisons: View previous close, week, month, and year data
  • Detailed Market Indicators: Access individual component scores including:
    • Market Momentum (S&P 500 & S&P 125)
    • Stock Price Strength & Breadth
    • Put/Call Options Ratio
    • Market Volatility (VIX)
    • Junk Bond Demand
    • Safe Haven Demand
  • Flexible Output: Choose between structured markdown or raw JSON format

Requirements

  • Node.js 18 or newer
  • VS Code, Cursor, Windsurf, Claude Desktop or any other MCP client

Getting Started

Local (Stdio)

First, install the Fear & Greed MCP server with your client. A typical configuration looks like this:

{
  "mcpServers": {
    "mcp-server-fear-greed": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-server-fear-greed@latest"
      ]
    }
  }
}
Install in VS Code

You can also install the mcp-server-fear-greed MCP server using the VS Code CLI:

# For VS Code
code --add-mcp '{"name":"mcp-server-fear-greed","command":"npx","args":["mcp-server-fear-greed@latest"]}'

After installation, the Fear & Greed MCP server will be available for use with your GitHub Copilot agent in VS Code.

Install in Cursor

Go to Cursor Settings -> MCP -> Add new MCP Server. Name to your liking, npx mcp-server-fear-greed. You can also verify config or add command like arguments via clicking Edit.

{
  "mcpServers": {
    "mcp-server-fear-greed": {
      "command": "npx",
      "args": [
        "mcp-server-fear-greed@latest"
      ]
    }
  }
}
Install in Windsurf

Follow Windsurf MCP documentation. Use following configuration:

{
  "mcpServers": {
    "mcp-server-fear-greed": {
      "command": "npx",
      "args": [
        "mcp-server-fear-greed@latest"
      ]
    }
  }
}
Install in Claude Desktop

Follow the MCP install guide, use following configuration:

{
  "mcpServers": {
    "mcp-server-fear-greed": {
      "command": "npx",
      "args": [
        "mcp-server-fear-greed@latest"
      ]
    }
  }
}
Remote (SSE / Streamable HTTP)

At the same time, use --port $your_port arg to start the browser mcp can be converted into SSE and Streamable HTTP Server.

# normal run remote mcp server
npx mcp-server-fear-greed --port 8089

You can use one of the two MCP Server remote endpoint:

  • Streamable HTTP(Recommended): http://127.0.0.1::8089/mcp
  • SSE: http://127.0.0.1::8089/sse

And then in MCP client config, set the url to the SSE endpoint:

{
  "mcpServers": {
    "mcp-server-fear-greed": {
      "url": "http://127.0.0.1::8089/sse"
    }
  }
}

url to the Streamable HTTP:

{
  "mcpServers": {
    "mcp-server-fear-greed": {
      "type": "streamable-http", // If there is MCP Client support
      "url": "http://127.0.0.1::8089/mcp"
    }
  }
}
In-memory call

If your MCP Client is developed based on JavaScript / TypeScript, you can directly use in-process calls to avoid requiring your users to install the command-line interface to use Fear & Greed MCP.

import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { InMemoryTransport } from '@modelcontextprotocol/sdk/inMemory.js';

// type: module project usage
import { createServer } from 'mcp-server-fear-greed';
// commonjs project usage
// const { createServer } = await import('mcp-server-fear-greed')

const client = new Client(
  {
    name: 'test fear greed client',
    version: '1.0',
  },
  {
    capabilities: {},
  },
);

const server = createServer();
const [clientTransport, serverTransport] = InMemoryTransport.createLinkedPair();

await Promise.all([
  client.connect(clientTransport),
  server.connect(serverTransport),
]);

// list tools
const result = await client.listTools();
console.log(result);

// call tool
const toolResult = await client.callTool({
  name: 'get_fear_greed_index',
  arguments: {
    format: 'json'
  },
});
console.log(toolResult);

API Reference

Tool: get_fear_greed_index

Fetches the current Fear & Greed Index and related market indicators.

Parameters
  • format (optional): Output format
    • "structured" (default): Returns formatted markdown with organized data
    • "json": Returns raw JSON data
Example Usage
// Get structured output
await client.callTool("get_fear_greed_index");

// Get JSON output
await client.callTool("get_fear_greed_index", { format: "json" });
Response Structure

The tool returns data in the following structure:

{
  "fear_and_greed": {
    "score": 75,
    "rating": "greed",
    "timestamp": "2025-07-18T23:59:57+00:00",
    "previous_close": 75.31,
    "previous_1_week": 75.26,
    "previous_1_month": 54.29,
    "previous_1_year": 45.94
  },
  "fear_and_greed_historical": {
    "timestamp": 1752883197000,
    "score": 75,
    "rating": "greed"
  },
  "market_momentum_sp500": {
    "timestamp": 1752871567000,
    "score": 61.2,
    "rating": "greed"
  },
  "market_momentum_sp125": {
    "timestamp": 1752871567000,
    "score": 61.2,
    "rating": "greed"
  },
  "stock_price_strength": {
    "timestamp": 1752883197000,
    "score": 80,
    "rating": "extreme greed"
  },
  "stock_price_breadth": {
    "timestamp": 1752883197000,
    "score": 84,
    "rating": "extreme greed"
  },
  "put_call_options": {
    "timestamp": 1752871897000,
    "score": 79.6,
    "rating": "extreme greed"
  },
  "market_volatility_vix": {
    "timestamp": 1752869701000,
    "score": 50,
    "rating": "neutral"
  },
  "market_volatility_vix_50": {
    "timestamp": 1752869701000,
    "score": 50,
    "rating": "neutral"
  },
  "junk_bond_demand": {
    "timestamp": 1752877800000,
    "score": 88.8,
    "rating": "extreme greed"
  },
  "safe_haven_demand": {
    "timestamp": 1752868799000,
    "score": 81.4,
    "rating": "extreme greed"
  }
}

Fear & Greed Index Ratings

The index uses the following rating scale:

  • 0-25: Extreme Fear
  • 26-45: Fear
  • 46-55: Neutral
  • 56-75: Greed
  • 76-100: Extreme Greed
Development

Access http://127.0.0.1:6274/:

npm run dev
Error Handling

The server includes comprehensive error handling:

  • Network request failures are caught and reported
  • Invalid API responses are handled gracefully
  • Missing data fields are filled with sensible defaults
  • All errors include descriptive messages

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.