What the MCP is
A server that lets an AI client use the platform's tools to work on your website.
MCP (Model Context Protocol) is the standard way for an AI client — Claude, Cursor, Codex… — to call a service's tools. The Webmati MCP server is at:
https://mcp.webmati.com/mcp
Connect it to your client (Connect), then give it work in plain words: "create a Contact page with a form", "add these 5 products from this file", "make an API /orders that saves orders to a table", "where did visitors come from this week?". The AI picks a tool, calls it, reads the result, and carries on.
What is in it
- 182 tools in groups: pages, theme, content, blog, media, translations, domains, the backend (tables, workflows, routes), apps, backups, analytics, selling — see Tools and skills.
- Skills: ready-made playbooks for big jobs (building a website, a section, a
backend, backups…). The server sends the list of skills as soon as a client
connects; the AI loads one with the
get_skilltool before it starts. - Each session works on one website: the key (or the sign-in) picks the workspace, and the website is chosen as described in Connect.
MCP or CLI?
The MCP suits working through a conversation, with nothing to install. When the job
grows — many theme files, Go code for workflows, git and review — move to the
CLI: tell the AI "I want to work locally" and it calls
init_local_project — see Work locally.