mcp-server-code-execution-mode — Claude Code skill: setup & use
mcp-server-code-execution-mode executes Python in rootless Podman or Docker containers and proxies MCP tools through a single entry point, cutting context from 30,000 to 200 tokens.
no license 173 installs ★ 9 updated 2026-10-09
npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a claude-codeWhat mcp-server-code-execution-mode does
This skill sets up Python execution inside a rootless Podman or Docker sandbox. Instead of loading hundreds of tool definitions directly into the agent context, it exposes a single run_python tool. The agent discovers connected MCP servers, inspects schemas, and invokes tools directly from script code.
The bridge supports fuzzy search across servers and keeps variables intact between steps. The sandbox runs with dropped Linux capabilities, a read-only filesystem, disabled network access, a default 512m memory limit, and a 128-process cap.
Without Podman or Docker set up in rootless mode, the bridge cannot start. To proxy additional MCP tools into the sandbox, define them in mcp_bridge_config.json via standard stdio commands.
What it helps with
- shrink context overhead when connecting dozens of MCP tools
- execute untrusted Python scripts in a network-isolated container
- discover and call MCP server tools dynamically at runtime
- maintain variables and session state across multiple execution turns
Example prompts for the agent
- Run this data analysis script inside the isolated Python container
- Search for GitHub MCP tools and create an issue through code execution
- Fetch calendar events and create corresponding tasks across proxied servers
How to install mcp-server-code-execution-mode
1Skills CLI (Vercel)
npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a claude-code2Manually: put the files in place
~/.claude/skills/mcp-server-code-execution-mode/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/mcp-server-code-execution-mode/
1Skills CLI (Vercel)
npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a codex2Manually: put the files in place
.agents/skills/mcp-server-code-execution-mode/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a cursor2Manually: put the files in place
.cursor/skills/mcp-server-code-execution-mode/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install reason-machines/mcp-skills skills/mcp-server-code-execution-mode2Manually: put the files in place
.github/skills/mcp-server-code-execution-mode/SKILL.md1Skills CLI (Vercel)
npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a gemini-cli2Platform CLI
gemini skills install https://github.com/reason-machines/mcp-skills --path skills/mcp-server-code-execution-modeManually: put the files in place
.devin/skills/mcp-server-code-execution-mode/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about mcp-server-code-execution-mode
How to install mcp-server-code-execution-mode in Claude Code?
Run npx skills add reason-machines/mcp-skills --skill mcp-server-code-execution-mode -a claude-code. The skill file will be created at ~/.claude/skills/mcp-server-code-execution-mode/SKILL.md.
What does mcp-server-code-execution-mode do?
It proxies MCP tools through Python execution, replacing individual tool definitions in the prompt with runtime discovery inside an isolated container.
What prerequisites are required before using it?
You need Podman or Docker running in rootless mode, the ghcr.io/elusznik/mcp-code-execution:latest image, and the mcp-code-execution Python package installed via pip or uv.