This plugin enables Claude Code to execute Python code in a local sandboxed Apple container using InstaVM CodeRunner via the Model Context Protocol (MCP).
# 1. Install and start CodeRunner (one-time setup)
git clone https://github.com/instavm/coderunner.git
cd coderunner
sudo ./install.sh
# 2. Install the Claude Code plugin
claude plugin marketplace add instavm/coderunner-plugin
claude plugin install instavm-coderunnerThat's it! Claude Code now has access to all CodeRunner tools.
IMPORTANT: You must have CodeRunner installed and running before installing this plugin.
git clone https://github.com/instavm/coderunner.git
cd coderunner
sudo ./install.shThis will:
- Install Apple's container runtime (version 0.8.0+)
- Pull the
instavm/coderunnercontainer image - Start the CodeRunner MCP service at
http://coderunner.local:8222/mcp
Verify CodeRunner is running:
curl http://coderunner.local:8222/execute -X POST -H "Content-Type: application/json" -d '{"code":"print(\"test\")"}'# Add the InstaVM marketplace from GitHub
claude plugin marketplace add instavm/coderunner-plugin
# Install the plugin
claude plugin install instavm-coderunner
$ claude
-
/plugin-->Marketplaces -
Add the InstaVM marketplace repository:
-
Install the
instavm-coderunnerplugin and reconnect with/mcp
Once installed, Claude Code can execute Python code in the sandboxed CodeRunner container. Simply ask Claude to run Python code, and it will automatically use the CodeRunner environment.
The plugin provides several capabilities:
- Execute arbitrary Python code with persistent state
- Access popular Python libraries (pandas, numpy, matplotlib, requests, etc.)
- Fetch and process web content with Playwright
- Work with files in a sandboxed environment
- List and use specialized skills for tasks like PDF manipulation, image processing, and more
- Sandboxed Execution: All code runs in an isolated Apple container
- Persistent Kernel: Jupyter kernel maintains state across executions
- Rich Libraries: Pre-installed data science and web scraping libraries
- Web Scraping: Built-in Playwright support for browser automation
- Specialized Skills: Access to pre-built skills for common tasks
- Secure: Network-isolated environment for safe code execution
The plugin exposes these MCP tools to Claude Code:
execute_python_code: Run Python code in a persistent Jupyter kernelnavigate_and_get_all_visible_text: Fetch and extract text from webpageslist_skills: View available specialized skillsget_skill_info: Get documentation for specific skillsget_skill_file: Access additional skill documentation files
-
Verify CodeRunner is running:
curl http://coderunner.local:8222/execute -X POST -H "Content-Type: application/json" -d '{"code":"print(\"test\")"}'
-
Reconnect MCP servers in Claude Code:
/mcp -
Check plugin installation:
claude plugin list
If CodeRunner fails to start:
cd coderunner
sudo ./install.shEnsure you have:
- macOS with Apple Silicon or Intel
- Administrator privileges for installation
- Apple container runtime (installed automatically by install.sh)
# Uninstall the plugin
claude plugin uninstall instavm-coderunner
# Remove the marketplace
claude plugin marketplace remove instavm/coderunner-plugin
# Stop CodeRunner (optional)
container stop coderunnerContributions are welcome! Please feel free to submit issues or pull requests to the GitHub repository.
This plugin is released under the MIT License.
- InstaVM CodeRunner - The sandboxed Python execution environment
- Claude Code - Anthropic's official CLI for Claude



