Manage your site by talking to AI
The MCP Server is a premium module that lets AI assistants like Claude and ChatGPT securely connect to your site and update content directly. It runs on the Model Context Protocol (MCP) standard, and every change it makes passes through role-based permissions and publish-approval.
How does it work?
MCP (Model Context Protocol) is an open standard that has emerged for connecting AI clients to applications. When you enable the module, your site becomes an MCP server: Claude Desktop connects with an API key and ChatGPT connects securely via OAuth 2.1. The connected assistant carries out the instructions you give in chat — "add an FAQ block to this page", "update this post's title", "redirect the old address to the new one" — through curated, validated tools.
The write surface is deliberately narrow: the AI is never handed a generic "change anything" power. It is offered only safe, individually defined tools such as editing page blocks, creating and updating documents, and adding redirects — so the outcome stays predictable and auditable.
Security: three independent gates
The AI's access to your site passes through three separate gates, and nothing happens unless all three are satisfied. First, the license: the module is entitled centrally. Second, your master switch: while the module is off in the panel no request is accepted, and the module ships OFF by default. Third, role-based permissions and publish-approval: the AI can never do more than the user it is connected to, and content it publishes drops into your approval flow when your rules require it.
Sensitive areas such as personal data, users, roles and licenses are never exposed to this surface. Where reading a sensitive field is allowed, personal data is masked — the AI gets its job done while the data stays protected.
MCP Server highlights
Claude & ChatGPT compatible
Supports both connection paths: an API key for Claude Desktop and OAuth 2.1 + PKCE for ChatGPT. Works seamlessly with standard MCP clients and needs no extra server or service.
Edit content by talking
Add and remove page blocks, reorder them, create and update posts and products, edit global settings and add redirects — all from chat, through individually defined tools.
Three-layer security
Every operation passes the license, the customer master switch and role-based permissions + publish-approval. No change is applied unless all three hold, and the module is off by default.
A curated, safe surface
The AI is not given a blanket "do anything" power; only narrow, validated tools are exposed. Deletions and sensitive fields are deliberately kept closed.
Integrated with publish-approval
Content created by the AI joins your existing publish-approval flow. Unauthorized publishing is blocked, and when needed content is held as a draft for human approval.
Sensitive-data protection & usage cap
Data under privacy law, users, roles and licenses are never exposed; personal data is masked in readable sensitive fields. A per-key daily request cap limits abuse.
How does the module work?
Each add-on module is an independent capability managed from one place with "master switch" logic. When you enable the module, the feature activates both in the admin panel and on the site's front end; when you disable it, it stops entirely. Modules are licensed and updated centrally — installation, versioning and maintenance are not your burden.