What is MCP, and how to deploy a site from an AI

MCP (Model Context Protocol) is an open protocol that lets AI models connect to external services safely and take actions through a set of tools. Put simply, it is a socket an AI agent plugs into to get things done in the real world, such as deploying sites.

Why a hosting platform needs it

Normally, to publish a site you open a control panel and click buttons. With MCP, the AI model does it for you: you write “build and deploy this project, attach the domain example.com”, and the agent calls the right tools itself.

How it works on Tuqo

1. Connect Tuqo to your model. Claude (claude.ai) and ChatGPT connect over OAuth: you paste only the address https://mcp.tuqo.ru/mcp, and sign-in and project selection happen on the Tuqo side, with no key to paste. CLI and IDE clients (Claude Code, Cursor, Gemini CLI, Qwen Code and others) use a project key tqk_… from the panel in the Authorization header:

   {
     "mcpServers": {
       "tuqo": {
         "url": "https://mcp.tuqo.ru/mcp",
         "headers": { "Authorization": "Bearer tqk_xxx_yyy" }
       }
     }
   }

   Step-by-step instructions for each client are on the Connect an AI page.

2. Just ask the model, and it calls the tools: create_site, deploy_files (publish static files as they are), deploy_site (build a project from source), set_custom_domain, activate_deploy (make a version live, or roll back), get_form_submissions (read form submissions) and more, 52 in total. The full list is on the MCP page.

What the agent can do

• Create and delete sites
• Deploy from source and follow the build status
• Attach and verify domains
• Roll back to earlier versions

Everything available in the panel is available to the agent. And when there is no AI model at hand, the same actions work through the REST API.

Connect your AI (Claude, ChatGPT, Cursor…) → · MCP docs → · API docs →