coderunner
A local sandbox for your AI agents
CodeRunner is an MCP server that provides a local, isolated sandbox environment for AI agents to safely execute Python code. It runs inside a container with VM-level isolation on macOS Apple Silicon, protecting your host system from potentially harmful agent actions. The server integrates with Claude Desktop, Claude Code, OpenCode, Gemini CLI, Kiro, and OpenAI agents, offering persistent Jupyter kernels, web scraping via Playwright, and a built-in skills system for common tasks like PDF manipulation and image processing.
Key Features
Use Cases
- 01Running untrusted AI-generated code safely without risking data loss or exfiltration
- 02Executing data science workflows with persistent kernel state across multiple agent interactions
- 03Automating PDF form filling and document processing tasks through AI agents
- 04Web scraping and data collection with isolated browser automation
- 05Testing AI agent behaviors in a clean, reproducible sandbox environment
- 06Running multiple AI agents concurrently with isolated execution contexts
Related MCP Servers
View moremarkitdown
Python tool for converting files and office documents to Markdown.
firecrawl
The context API to search, scrape, and interact with the web at scale. 🔥
prompts.chat
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.
langflow
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
coderunner — FAQ
What is CodeRunner and what does it do?+
CodeRunner is an MCP server that provides a sandboxed execution environment for AI agents to run Python code safely. It uses VM-level container isolation to protect your host system from potentially harmful agent actions while offering persistent Jupyter kernels, web scraping, and built-in skills for common tasks.
How do I install and start CodeRunner?+
Clone the repository, run the install.sh script on a Mac with Apple Silicon (M1/M2/M3/M4), and it will set up the sandbox container automatically. The installer requires macOS, Apple Silicon, and Python 3.10 or higher.
Which AI clients work with CodeRunner?+
CodeRunner integrates with Claude Desktop, Claude Code CLI, OpenCode, Gemini CLI, Kiro by Amazon, and OpenAI agents via Python. Each client connects to the MCP server at http://coderunner.local:8222/mcp or through a local proxy.
Do I need an API key to use CodeRunner?+
CodeRunner itself does not require an API key. However, the AI clients you connect to it (like OpenAI agents) will need their respective API keys configured separately.
Is CodeRunner free and open source?+
Yes, CodeRunner is open source and licensed under Apache 2.0, making it free to use, modify, and distribute.
Can I disable network access from the sandbox?+
Yes, you can install CodeRunner with network isolation by running 'CODERUNNER_NETWORK=none ./install.sh', which restricts the sandbox to host-only networking with no internet access.
How do I install coderunner?+
Open the source repository on GitHub and follow its README. coderunner is a mcp server — MCP Agents Market links you directly to the official repo.
Is coderunner free?+
coderunner is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.