Code Execution Mode

by elusznik · File Systems · mcp-server, python, ai

Execute Python in isolated containers while bridging access to other MCP servers through auto-discovered proxies

MCP server providing secure Python code execution in isolated containers while enabling seamless integration with other MCP servers through bridge architecture. Automatically discovers MCP servers from Claude Code and standard configuration locations, then executes Python code in rootless containers (Podman/Docker) with network isolation and resource limits while providing runtime access to discovered MCP servers through generated proxy objects. Features TOON-format response encoding for token efficiency, persistent MCP client sessions, automatic server metadata discovery with tool aliasing, and comprehensive sandbox security including memory limits, process restrictions, and temporary filesystem isolation.

Source: https://github.com/elusznik/mcp-server-code-execution-mode

Install

git clone https://github.com/elusznik/mcp-server-code-execution-mode

Tags: mcp-server, python, ai

⭐ 319 GitHub stars · Source: pulsemcp

About File Systems MCP servers and Claude skills

File Systems MCP servers extend what AI agents can do inside Claude Code, Cursor, Copilot, Codex, and Windsurf. The Skiln directory indexes 16,000+ such integrations across 22 categories.

Code Execution Mode is one of hundreds of File Systems entries indexed on Skiln. Browse the full File Systems category or the complete directory of Claude skills, MCP servers, agents, commands, and hooks.

Related File Systems MCPs and skills

  • Rust Docs by govcraft

    Efficiently loads and processes Rust crate documentation using LlamaIndex's HTML reader, with intelligent file selection logic to handle duplicate files and optional comprehensive parsing modes.

  • MATLAB by matlab

    Run MATLAB using AI applications with the official MCP server from MathWorks, enabling code execution, evaluation, and static analysis

  • Dify by yanxingliu

    Integrates with the Dify API to enable workflow execution and management for automated task processing and decision making across domains.

  • Apollo GraphQL by apollographql

    Enables AI agents to interact with GraphQL APIs by exposing operations as tools, supporting both local schema files and Apollo GraphOS integration with features for introspection and custom scalar mapping.

  • Pipelock by luckypipewrench

    Security harness that wraps MCP servers as stdio proxies with credential exfiltration prevention, DLP pattern matching, and SSRF protection.

  • LLMBasedOS by iluxu

    Secure Arch Linux gateway that bridges LLMs with local system capabilities through specialized servers for file system, email, sync, and agent operations without exposing sensitive information

  • Strava by r-huijts

    Integrates with Strava's API to access and analyze fitness data, retrieve athlete profiles, activities, segments, routes, and clubs, with features for exporting routes in GPX/TCX formats for training insights and route planning.

  • Garmin Connect by taxuspt

    Integrates with Garmin Connect to expose fitness and health data, enabling access to activities, detailed workout information, and health metrics like heart rate through natural language interactions.

Frequently asked questions

How do I install Code Execution Mode?

Add the install command above to your Claude Code, Cursor, or Windsurf MCP configuration. Most servers register via npx, a local command, or a Docker image. Refer to the source repository for environment variables and credential requirements.

Which clients support Code Execution Mode?

Any MCP-compatible client works: Claude Desktop, Claude Code CLI, Cursor, Windsurf, Zed, and VS Code with the official MCP extension. OpenAI Codex and GitHub Copilot increasingly support MCP via adapter bridges.

Is Code Execution Mode free?

The server itself is typically open source. Any upstream service (API keys, paid tiers, hosted infrastructure) may have its own pricing. Check the source repository for details.