WordPress now powers over 43% of the web, but in 2025, it’s evolving into something bigger, an AI operations hub. If you’re an agency managing dozens of sites, you’ve probably asked: “How can I integrate AI into WordPress without relying on 10 different tools?”
The answer is the WordPress MCP Server. It connects large language models like Claude directly to your WordPress dashboard, turning natural language prompts into secure, automated actions.
Instead of clicking through wp-admin, you can simply say: “Publish this post,” “Update plugins,” or “List pending comments.” The WordPress MCP server for agencies makes routine work faster, scalable, and client-ready.
Table of Contents
Quick answer: What is WordPress MCP server and how do you use it?
A WordPress MCP server is a backend service that lets AI agents manage WordPress using Model Context Protocol.
It converts natural language prompts into authenticated REST API calls that create posts, manage plugins, users and more.
Run the server, connect it to your WordPress sites with application passwords, then connect an AI client like Claude Desktop and start issuing prompts.
What Is the WordPress MCP Server?
The WordPress MCP Server, also known as the MCP for WordPress, is a backend layer that connects large language models (LLMs) like Claude directly with your WordPress sites.
Instead of logging into multiple wp-admin dashboards or installing dozens of plugins, the MCP server acts as a translator: it takes your natural language prompts and converts them into secure API requests that WordPress understands. This means you can perform real tasks, publishing posts, moderating comments, and updating plugins, by simply asking your AI tool in plain English.
For example, an agency developer could type or say:
- Create a new draft blog post titled ‘Holiday Campaign Ideas.
- Update WooCommerce plugins on all client stores.
- List comments awaiting moderation on the travel_blog site.
The MCP server processes these instructions and executes them instantly, without a single manual click.
For instance, InstaWP’s WordPress MCP server can create a post without dropping a drop of sweat.
How to Set Up the WordPress MCP Server
Getting started with the WordPress MCP server is easier than you think. You don’t need custom infrastructure or advanced DevOps knowledge.
In fact, with InstaWP’s MCP server, you can be up and running in minutes, using just your terminal and a WordPress instance.
The InstaWP’s WordPress MCP Server is an open-source, developer-first implementation of the Model Context Protocol (MCP) designed to bridge AI agents like Claude with your WordPress sites. Built using Node.js and TypeScript, it translates natural language prompts into secure, authenticated API actions, such as creating posts, managing plugins, updating users, and more.
Unlike generic MCP setups, InstaWP’s server is modular, production-ready, and tailored for WordPress. It comes with prebuilt tools for core operations (posts, pages, media, users, comments, categories, and plugins), environment-based configuration, and instant compatibility with Claude Desktop via a sample config file.
- WordPress-Native Tooling: Unlike generic MCP projects, InstaWP’s version includes tools specifically for WordPress functionality; everything from media uploads to plugin management is already mapped.
- Plug-and-Play with Claude: With a ready-to-use Claude Desktop config, integration is seamless—no complex setup needed.
- One-Line Launch via npx: Instantly spin up the server using
npx -y @instawp/mcp-wp—no install or repo cloning required. - Safe Staging with InstaWP: Test and deploy with disposable sandboxes, application passwords, and zero impact on live sites.
For agencies, this means faster deployment, safer experimentation, and a head start on AI-native WordPress automation—without writing middleware from scratch.
What You’ll Need Before You Start
To set up your AI-first WordPress workflow, you need:
- Node.js (version 18 or higher)
- A WordPress site (staging or live) with REST API access
- Application password (for secure API auth)
- Claude Desktop (or any compatible LLM tool)
Here is how you can set up this WordPress MCP server to integrate AI in WordPress.
Step 1: Install the MCP Server
Run this in your terminal:
npx -y @instawp/mcp-wpAlternatively, clone the GitHub repo and install dependencies manually:
git clone <repo-url>
cd wordpress-mcp-server
npm installStep 2: Create Your .env File
In the root of your MCP folder, create a .env file:
WORDPRESS_API_URL=https://your-site.com
WORDPRESS_USERNAME=your_username
WORDPRESS_PASSWORD=your_app_password
💡 Tip: If you’re using InstaWP’s WordPress sandboxes, just copy your site URL and credentials from the Login Credential Viewer.
Step 3: Configure Claude Desktop
- Open Claude Desktop > Settings > Developer tab
- Open and edit
claude_desktop_config.json - Copy the config from
claude_desktop_config.json.examplein the repo - Update it with the absolute path to your MCP server and the WordPress environment variables
- Save and restart Claude Desktop
Now, Claude will know how to communicate with your MCP Server for WordPress.
Step 4: Run the Server
You can now run the server in development or production:
npm run dev
# or
npm start
Claude will now interpret your natural language prompts and relay them to WordPress through this server.
Set up WordPress’s Official MCP Server
While InstaWP’s MCP Server is one of the most powerful and developer-friendly implementations available today, we understand that some agencies—especially those with highly customized stacks or self-hosted DevOps pipelines—may prefer to work directly with the official WordPress MCP Server.
Below, we’ll walk you through how to install and configure the official WordPress MCP Server from scratch.
Step 1: Install the WordPress MCP Server
The official WordPress MCP server is open-source and published on npm. You can install it on your local machine, a cloud VM, or even containerize it for broader deployments.
npm install server-wp-mcpThat’s it. You’ve just taken the first step toward AI-first WordPress management.
This package acts as the bridge between your sites and the AI models. It’s the beating heart of every MCP for WordPress setup. And if you’re working across multiple projects, you can run one server to manage them all—centralized, scalable, and elegant.
Step 2: Generate WordPress Application Passwords
Now that the WordPress MCP server is installed, you need a secure way for it to access your WordPress sites. This is where application passwords come in.
Go to your client’s WordPress dashboard and follow these steps:
- Navigate to Users → Profile
- Scroll to Application Passwords
- Add a label (e.g., “MCP Server Access”)
- Click Add New Application Password
- Copy the generated password securely

Do this for each site you want to connect. This is how the WordPress MCP server will securely authenticate API calls without exposing the main login.
Bonus Tip: WordPress staging environments make this setup even safer. Test your MCP integrations in a disposable WordPress sandbox before going live.
Step 3: Create Your Server Configuration
Now that you’ve installed the WordPress MCP server and created your secure application passwords, it’s time to tell the server which WordPress sites it should connect to. You’ll do this using a simple file called wp-sites.json.
Wondering what it is? It’s just a small text file (in JSON format) that lists the WordPress sites you want to control using AI, along with their login credentials (application passwords, not your main passwords!).
This is how you can create one.
- Open a text editor (like Notepad, VS Code, or Sublime). InstaWP comes with a VS Code editor built into the staging sites.
- Copy and paste the following structure:
{
"client_one": {
"URL": "https://clientone.com",
"USER": "admin",
"PASS": "your-app-password"
},
"client_two": {
"URL": "https://clienttwo.com",
"USER": "admin2",
"PASS": "another-password"
}
}
- Replace:
- clientone.com and clienttwo.com with your actual site URLs.
- admin and admin2 with the real usernames used for logging into WordPress.
- your-app-password with the application password you generated from Step 2 for each site.
- clientone.com and clienttwo.com with your actual site URLs.
- Save this file with the exact name wp-sites.json.
Let’s say you run an agency that manages two sites:
- A blog site for a personal brand
- An e-commerce site for a client
You could name your entries like this:
{
"personal_blog": {
"URL": "https://myblog.com",
"USER": "john_doe",
"PASS": "abc123appPassword"
},
"client_shop": {
"URL": "https://shopclient.com",
"USER": "admin_shop",
"PASS": "xyz456appPassword"
}
}Now, your AI model knows how to reach both.
This configuration file is the central address book for your WordPress MCP server. It tells the server:
- Where your WordPress sites live (their URLs)
- How to log in securely
- What names to use when sending instructions
Whether you’re managing 2, 20, or 200 sites, this approach scales cleanly and securely.
Step 4: Launch Your WordPress MCP Server
Okay, you’ve installed the WordPress MCP server and created your wp-sites.json file (which lists the sites your AI will connect to). Now it’s time to actually launch the server so it can start talking to your WordPress sites.
1. Set the environment variable
Copy and paste this into your terminal:
export WP_SITES_PATH=~/path/to/wp-sites.jsonIt tells the WordPress MCP server where to find your wp-sites.json config file.
Replace this part:~/path/to/wp-sites.json with the actual location of the file on your computer.
For example, if you saved it on your desktop:
export WP_SITES_PATH=~/Desktop/wp-sites.json
If you’re on Windows using PowerShell, you’ll write it like this:
$env:WP_SITES_PATH = "C:\Users\YourName\Desktop\wp-sites.json"
2. Start the Server
Now run:
npm startThis command actually starts the WordPress MCP server, so it can begin listening for instructions from AI agents like Claude.
If everything is working, your terminal will show something like:
WordPress MCP server started with 2 site(s) configured
This means your MCP server has read the config file and is ready to communicate with the WordPress sites listed inside it. Success!
Now that your WordPress MCP server is running, you can test it by connecting Claude (or any compatible LLM). Try prompts like:
“Show me the last 5 posts from client_one”
“Create a new draft post on client_two titled ‘Welcome to AI-Powered WordPress’”
“List all plugins active on client_one”
The AI model sends those commands to your MCP for WordPress setup, and your WordPress MCP server turns them into API calls your site understands.
Step 5: Test Your WordPress MCP Setup with Claude
Once your WordPress MCP server is running, connect it to Claude or another AI tool that supports JSON tool calling. You can now issue real-time commands like:
- “Fetch all draft posts from client_one”
- “Publish a new post titled ‘AI in WordPress 2025’ on client_two.”
- “Get plugin version list for client_one”
These instructions go through the WordPress MCP server, which parses them into authenticated API requests. It’s clean, efficient, and game-changing for agencies.
Pro Tip for Agencies: You don’t need to build this from scratch. InstaWP provides a production-ready WordPress MCP Server implementation with WordPress-native tooling, a sample Claude Desktop configuration, and safe staging environments. You can test workflows in disposable sandboxes, then roll them out across client sites confidently, without writing custom middleware.
Why Should Agencies Care About WordPress MCP?
Agencies spend countless hours on routine WordPress site management; approving updates, fixing plugin issues, publishing client content, cleaning up spam, and keeping dashboards alive. None of this is what you built your agency for. It’s the digital equivalent of babysitting dozens of sites at once.
The WordPress MCP Server, sometimes called the MCP for WordPress, changes that reality. By linking your sites to Claude or any LLM, you can offload nearly half of those repetitive tasks to AI.
If you manage 20+ WordPress sites, your team probably repeats the same tasks every week:
- Updating plugins and themes
- Publishing blog posts or drafts
- Running SEO and performance checks
- Cleaning up spam comments
- Monitoring uptime
Now, imagine delegating all of that to Claude or another AI model, using one shared MCP server for WordPress to deliver: AI-powered content engines, automated performance alerts, real-time client dashboards, and on-demand WordPress assistants.
Just as pairing AI tools with a virtual administrative assistant creates a hybrid workflow for domains like IT, HR, marketing, and so on, WordPress developers are building AI-powered website assistants for their clients using the WordPress MCP protocol. These assistants pull analytics, draft content, and even generate WordPress maintenance reports, bringing automation directly into everyday website management.
The MCP protocol does more than ping you with reminders. It executes actions:
- “Show me all draft posts from client_blog_1.”
- “Publish the top 3 articles Claude just wrote for realestate_site.”
- “Update all plugins on our WooCommerce stores.”
Your AI assistant becomes a full-service site manager.
The Bigger Picture: From Sites to Services
We’re entering an era where agencies won’t just offer “WordPress development” but will position themselves as AI-enhanced digital partners. With the right tools, agencies are becoming a smart platform provider.
The WordPress MCP server helps you cross that line.
You’re not just building websites anymore. You’re delivering:
- AI-powered content engines
- Automated performance alerts
- Real-time client dashboards
- On-demand WordPress assistants
These are not ideas for the future. Agencies are already offering them today
How to Use WordPress MCP to Monetize AI Workflows
AI in WordPress isn’t just a productivity upgrade. It’s a business model.
If you’re an agency, freelancer, or product creator, the WordPress MCP server gives you a direct path to monetize automation, without needing to build your own AI platform.
With Claude + WordPress + InstaWP’s WordPress MCP, you can sell real services, launch scalable SaaS products, or bundle prebuilt solutions into monthly retainers.
1. Sell “AI Admin Assistant” Plans to Clients
Offer clients a virtual WordPress assistant that handles:
- Blog publishing
- SEO checks
- Plugin audits
- Spam moderation
- Basic reporting
Position it as:
“Your WordPress Admin Assistant—powered by Claude, available 24/7.”
Charge: $50–$200/month per site
Run it using: InstaWP MCP Server + Claude Desktop + InstaWP’s Performance Scanner & Activity Log Viewer
2. Launch Your Own AI-Powered WaaS or SaaS
Use InstaWP templates + MCP tooling to build an AI-driven platform:
- Create demo templates with Claude prompts preloaded
- Offer clients tiered plans based on AI access (e.g., content only vs full control)
- Deploy sites instantly using InstaWP WaaS capabilities
Examples:
✔️ “Let AI manage your content calendar”
✔️ “SEO checker powered by Claude”
✔️ “AI blog package for WooCommerce store owners”
3. Productize Your Workflows as Templates and Kits
Every agency has repeatable workflows. Why not sell them?
- Build Claude-compatible prompt kits
- Package InstaWP site templates with pre-installed MCP servers
- Offer “Done-for-You” AI assistant bundles on your own site or marketplaces
Example:
“AI Assistant for Bloggers – Includes SEO prompts, content calendar, and Claude integration.”
4. Add AI Consulting to Your Retainer Packages
Most clients don’t know what’s possible with AI.
Offer:
- WordPress AI workflow audits
- Claude + MCP onboarding
- Custom prompt design
- Ongoing support
Position yourself as more than a developer—you’re an AI integration partner.
Want even more control? You can extend the WordPress MCP server by building your own custom tools.
This gives your agency the power to productize these AI interactions into services.
Security Best Practices for Live MCP Usage
As powerful as the WordPress MCP server is, it must be handled responsibly, especially in live production environments. Here’s what we recommend:

And yes, MCP for WordPress has built-in structures to handle these concerns if you configure it thoughtfully.
Using WordPress staging sites means you can fully test your WordPress MCP server in isolated environments—no risk to production, full visibility into how tools behave.
Final Thoughts: Be the First, Be the Fastest
WordPress MCP isn’t just a tool. It’s a shift in how agencies work.
You can:
- Save 100s of hours per year through automation
- Offer AI assistants as a service
- Productize workflows with a hosted WordPress MCP server
- Scale client management across 10, 100, or 1000 sites
- Become the agency that leads with AI—not the one playing catch-up
The MCP for WordPress puts your agency at the forefront of this change. And InstaWP ensures you can experiment, deploy, and scale with confidence.
Start Building AI-First WordPress Sites with InstaWP Today
Want to get hands-on with the WordPress MCP server, without worrying about setting up servers or breaking a live site?
👉 Create a free InstaWP sandbox
Spin up a WordPress site in seconds.
Test Claude + MCP integrations safely.
Clone, sell, and deploy your AI-powered templates.
All without touching cPanel.
Whether you’re a solo dev or scaling agency, MCP for WordPress + InstaWP is your new superpower.
FAQs
Q1: Do I need to know advanced coding to use WordPress MCP?
Nope! If you’re comfortable editing simple JSON and running terminal commands, you’re good to go.
Q2: Is the WordPress MCP server safe for production use?
Yes—just follow security best practices like using HTTPS, app passwords, and read-only scopes where needed.
Q3: Can I use MCP for multiple clients?
Absolutely. You can add 10, 100, or 1000 sites to your config and manage them all through a single WordPress MCP server.
Q4: Does InstaWP support MCP workflows?
Yes! InstaWP is the perfect staging + testing platform for MCP for WordPress workflows—plus you can turn those workflows into sellable templates.
Q5: Can I sell this as a SaaS product?
Definitely. Package your WordPress MCP server and tools as a hosted service, and let clients connect their sites via an API key.