Purple Flea MCP Servers
for Claude

What Is Model Context Protocol?

Model Context Protocol (MCP) is an open standard developed by Anthropic that allows AI models like Claude to securely call external tools, read live data sources, and execute real-world actions during a conversation. Think of it as a universal plugin layer: instead of baking tool logic into every prompt, you register a server once and Claude automatically discovers its capabilities.

MCP servers communicate with Claude over a lightweight JSON-RPC transport — either stdio (for Claude Desktop) or HTTP with Server-Sent Events (for the Claude API). When Claude determines a tool call is appropriate, it sends a structured request to the MCP server, receives a structured response, and incorporates the result naturally into its reply. The user never has to write a single line of function-calling code.

Purple Flea's MCP server is purpose-built for on-chain and financial tooling. It packages nineteen production-ready tools across six categories — casino, trading, wallets, domains, faucet, and escrow — into a single npm package that Claude can use out of the box.

All 19 MCP Tools at a Glance

Every tool is fully typed with JSON Schema, supports structured error responses, and is safe to call concurrently. Claude will automatically choose the right tool based on your conversation — no explicit invocation syntax needed.

Casino Tools

Trading Tools

Wallet Tools

Domain Tools

Faucet Tools

Escrow Tools

Set Up in Under 5 Minutes

Purple Flea's MCP server ships as a standard npm package. You install it globally, drop a config block into your Claude Desktop settings file, and restart the app. Claude will immediately list the tools in its system context and use them whenever relevant.

npm install -g @purpleflea/mcp-server

purpleflea auth login --key YOUR_API_KEY

{
  "mcpServers": {
    "purpleflea": {
      "command": "purpleflea-mcp",
      "args": [],
      "env": {
        "PURPLEFLEA_API_KEY": "YOUR_API_KEY_HERE",
        "PURPLEFLEA_WALLET_ID": "YOUR_WALLET_ID",
        "PURPLEFLEA_NETWORK": "mainnet"
      }
    }
  }
}

How Claude Uses the Tools in a Conversation

Below is a real exchange showing Claude autonomously chaining multiple MCP tool calls — checking a balance, executing a token swap, opening a trading position, and then confirming the result. No special syntax is required from the user; Claude infers tool intent from natural language.

{ "chain": "ethereum", "asset": "ETH" }
// → { "balance": "0.847 ETH", "usd_value": "$2,114.50" }

{ "from": "ETH", "to": "USDC", "amount": "0.1", "chain": "ethereum" }
// → { "received": "249.82 USDC", "tx_hash": "0xabc...def", "route": "1inch" }

{ "market": "BTC-PERP", "side": "long", "collateral": "249.82 USDC", "leverage": 2 }
// → { "position_id": "pos_7x92k", "size": "0.0098 BTC", "entry_price": "$50,981" }

Works With the Claude API via HTTP Transport

Claude Desktop uses stdio transport, but if you're building an application on top of the Claude API you can connect Purple Flea's MCP server over HTTP with Server-Sent Events. This lets your backend relay tool calls to the Purple Flea server without the user needing Claude Desktop installed at all.

Start the Purple Flea MCP server in HTTP mode on any port, then pass the server URL when you create a Claude API session. The API will discover available tools from the /mcp/tools endpoint and route tool-use blocks to your server automatically.

purpleflea-mcp --transport http --port 3099 --api-key YOUR_KEY

import Anthropic from '@anthropic-ai/sdk';

const client = new Anthropic();

const response = await client.beta.messages.create({
  model: 'claude-opus-4-6',
  max_tokens: 1024,
  mcp_servers: [
    {
      type: 'url',
      url: 'http://localhost:3099/mcp',
      name: 'purpleflea'
    }
  ],
  messages: [
    { role: 'user', content: 'Flip a coin and bet 10 USDC on heads' }
  ]
});

console.log(response.content);

More Ways to Use Purple Flea

Purple Flea's APIs power agents across every major framework. Explore integrations and deep-dive API references below.