website-publisher
AI Website builder and publisher MCP. Quickly publish and deploy your AI generated code as real website URL. Support html, css, js, python etc.
claude mcp add --transport stdio 11cafe-website-publisher-mcp npx -y code-sandbox-mcp@latest
How to use
Runbox Website Publisher MCP provides a remote coding environment that can publish AI-generated website code as a real live website accessible via a URL. It offers a live code editor to view and modify the generated website files (HTML, CSS, JavaScript) and supports building, reading, and writing code inside a remote sandbox. You can also safely execute commands in the remote sandbox to test and fix issues, and publish the final site so it serves as a real website URL for demonstrations or sharing. To use it, simply request a website from Claude or the assistant (e.g., “Create a XYZ website and publish to Runbox”), and the MCP will spin up a publishable site and provide a URL such as https://runbox.ai/site/<some-id> where your hosted site can be viewed. The system also includes a live editor for iterating on code in real time and re-publishing updates to the live URL. The developer reference points to a server backend at https://github.com/11cafe/code-sandbox-server which powers the remote sandbox and publishing workflow. The MCP package you’ll run is code-sandbox-mcp, which orchestrates this capability via npx.
The workflow typically involves generating the site content, opening the remote sandbox to edit or refine code, then building and publishing the site to a public domain or runbox URL for easy sharing. The capabilities cover static site publishing (HTML, CSS, JS), multi-file projects, on-demand live editing, and command execution within a sandbox to test and debug before going live.
How to install
Prerequisites:
- Node.js and npm installed on your workstation (npx is included with npm 5.2+).
- Internet access to fetch the MCP package from npm.
Installation steps:
-
Verify Node.js and npm are installed: node -v npm -v
-
Prepare the MCP configuration (optional but recommended): create a file mcp_config.json with the following content: { "mcpServers": { "runbox-website-publisher": { "command": "npx", "args": ["-y", "code-sandbox-mcp@latest"] } } }
-
Run the MCP server using the provided npm package via npx: npx -y code-sandbox-mcp@latest
or, if you want to point to a specific configuration file:
npx -y code-sandbox-mcp@latest --config mcp_config.json
-
Follow the prompts or monitor logs for the creation of a live website publish session. Once ready, you’ll receive a URL to access the published site and the live editor.
Notes:
- No additional global installation is required beyond Node.js/npm unless you prefer a local setup; the MCP is run via npx for a transient, dependency-managed experience.
- If your environment blocks npm/yarn binary fetches, ensure network access to npm registry (https://registry.npmjs.org).
Additional notes
Tips and common considerations:
- The MCP publishes AI-generated code as a real URL; ensure you have the rights to publish any generated content and any assets you include.
- The live editor supports multi-file projects (HTML, CSS, JS and other assets). Use the editor to iterate on content and push updates to the live site.
- The developer server for the underlying sandbox is at https://github.com/11cafe/code-sandbox-server; expect remote sandbox restraints and safety checks during command execution.
- If you experience delays publishing or URL generation, check the network access from the hosting environment and verify that the code-sandbox-mcp package is up to date.
- Environment variables can be extended in the MCP runtime if needed (e.g., for custom domains or API keys), but the default configuration provided relies on the standard published workflow.
Related MCP Servers
any-chat-completions
MCP Server for using any LLM as a Tool
template
Template to quickly set up your own MCP server
n8n-workflow-builder
MCP server that allow LLM in agent mode builds n8n workflows for you
openai -agent-dotnet
Sample to create an AI Agent using OpenAI models with any MCP server running on Azure Container Apps
finance
LLM-powered MCP server for building financial deep-research agents, integrating web search, Crawl4AI scraping, and entity extraction into composable analysis flows.
unity-editor
An MCP server and client for LLMs to interact with Unity Projects