mcp-server-code-execution-mode

Overview

This MCP server implements Anthropic's and Cloudflare's ideas for reducing MCP tool definitions context bloat. Instead of exposing hundreds of individual tools to the LLM, it exposes one 'run_python' tool where the LLM writes Python code to discover, call, and compose other tools. This reduces context from 30K tokens to 200 tokens while proxying any stdio server. The server runs code in rootless Docker/Podman containers with strict security isolation (no network access, read-only filesystem, dropped capabilities).

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• run_python

  Executes Python code in an isolated container environment with MCP server proxying capabilities.

  Try:Analyze this data using pandasTry:Generate a chart from the sales dataTry:Search the GitHub API for recent issues

Comparable tools

Installation

{
  "mcpServers": {
    "code-execution": {
      "command": "python",
      "args": ["-m", "mcp_server_code_execution_mode"],
      "env": {}
    }
  }
}

FAQ

What security measures does this server implement?

The server runs code in rootless containers with network isolation, read-only filesystem, dropped capabilities, unprivileged user, and resource limits.

How does this reduce context usage compared to traditional MCP?

Instead of loading all tool schemas into context (30K+ tokens), it exposes only one 'run_python' tool and loads schemas dynamically when requested (200 tokens).

On Hacker News

Recent discussion from the developer community.

• MCP Code Execution Server: Zero-Context Discovery for 100 MCP Tools▲ 2
  Story by rahimnathwani · 2025-12-19

Compare mcp-server-code-execution-mode with