What a WordPress MCP server is
MCP, the Model Context Protocol, is the open standard AI tools use to talk to other software. WebCrew runs one hosted MCP endpoint for all your connected WordPress sites. Instead of giving an AI tool an application password to each site, you connect it once to WebCrew and choose what it may see and do.
Which tools
Any MCP client. We test with Claude, ChatGPT, Cursor, GitHub Copilot, Gemini, Perplexity, Mistral's Le Chat, Windsurf, Raycast, JetBrains AI, VS Code and n8n. Product names belong to their owners; WebCrew is not affiliated with them.
What you can ask
- "Which of my sites has updates waiting?"
- "Show me the latest check for harbourview.nl."
- "Which sites still run PHP 8.1?"
- "Which plugins have a known vulnerability right now, and on which sites?"
- "Draft new opening hours on bakkerij-vos.nl's contact page." The draft waits for approval in wp-admin like any other.
How to connect
- In the WebCrew dashboard, open Your own AI and create a connection. You get one address.
- In your AI tool, add a connector, integration or MCP server and paste the address. Claude, ChatGPT, Cursor and VS Code each have a menu for this.
- Sign in once with your WebCrew account (OAuth) and choose the scopes. Nothing is exposed by default.
Scopes: reading and writing are separate
A tool with read scope can see status, versions, findings and history and cannot change a site. A tool with write scope can create drafts, and every draft goes through the same approval loop as a request made in wp-admin: restore point, difference, a person's go. There is no scope that writes to a live site directly.
| Scope | Allows | Never allows |
|---|---|---|
| Read: sites | List sites, health scores, uptime, versions, findings | Any change |
| Read: content | Read posts, pages, products, fields on chosen sites | Any change |
| Write: drafts | Create a draft that waits for approval | Publishing without a person's go |
What it costs
Connecting your own AI is included on Free and Standard. Questions and drafts draw from the same token balance the assistant uses: 1,000 starter tokens, then €5 per 1,000.