home / mcp / toon-parse mcp server

toon-parse MCP Server

Provides a token-efficient MCP server that converts data to TOON and optimizes code context for AI agents.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "ankitpal181-toon-parse-mcp": {
      "command": "python3",
      "args": [
        "-m",
        "toon_parse_mcp.server"
      ]
    }
  }
}

You can deploy the toon-parse MCP Server to optimize token usage by converting data to the compact TOON format and removing non-essential context from code files, making AI agents work more efficiently with code and data inputs.

How to use

Start the toon-parse MCP Server and connect it from your MCP client. Once active, your client gains access to two tools that optimize data and code contexts: optimize_input_context for transforming raw inputs (JSON, XML, YAML, CSV) into TOON, and read_and_optimize_file for condensing local code files by removing inline comments and unnecessary whitespace while preserving useful structure. You can also rely on the built-in efficiency protocol to guide token-saving behavior.

How to install

Prerequisites before installation include having Python 3.10 or newer installed on your system. The server relies on Python packages and standard MCP tooling.

Install the toon-parse MCP Server package using Python's package manager:

pip install toon-parse-mcp

Configuration

To run the server locally, start the module using Python so it serves as an MCP endpoint that your clients can connect to.

Typical runtime command to start the server:

python3 -m toon_parse_mcp.server

Additional setup for MCP clients

Configure your MCP clients to recognize toon-parse as a server. Use the following example configurations in the client’s MCP settings to register toon-parse-mcp as an active MCP server.

{
  "mcpServers": {
    "toon-parse-mcp": {
      "command": "python3",
      "args": ["-m", "toon_parse_mcp.server"]
    }
  }
}

Available tools

optimize_input_context

Processes raw inputs written in JSON, XML, YAML, or CSV and returns a token-optimized TOON representation.

read_and_optimize_file

Reads a local code file and returns a token-optimized version by removing inline comments and minimizing whitespace while preserving functionality.