xron-mcp: Model Context Protocol (MCP) Compression Proxy

A lightweight MCP middleware that wraps any existing MCP server and automatically compresses JSON tool responses using XRON before they reach the LLM. This reduces token consumption by 60-80% for data-heavy tool outputs like BigQuery queries or Google Ads API results.

Features

• Stateless Proxy: Intercepts tools/call and tools/list transparently.
• Auto-Compression: Detects JSON text in tool responses and applies XRON's optimal compression.
• LLM-Friendly: Adds a format hint to help the model interpret the compressed data.
• No Configuration: Wraps standard MCP servers with a simple --wrap command.

Installation

npm install -g xron-mcp

Usage in mcp_config.json

Wrap any existing MCP server using xron-mcp --wrap:

{
  "mcpServers": {
    "bigquery-compressed": {
      "command": "xron-mcp",
      "args": ["--wrap", "npx", "-y", "@ergut/mcp-bigquery-server"],
      "env": {
        "GOOGLE_CLOUD_PROJECT": "your-project-id"
      }
    }
  }
}

How It Works

1. LLM/Agent calls a tool through xron-mcp.
2. xron-mcp forwards the call to the Upstream MCP Server.
3. The server returns its response.
4. xron-mcp scans the response content:
   • If it's plain text/JSON, it applies XRON.stringify(data, { level: 'auto' }).
   • If it's an image/binary, it passes it through unchanged.
5. The compressed response is returned to the LLM, saving 60-80% on the response's context window tax.

License

MIT