Massblogger MCP Server

Quick start — hosted endpoint

The fastest way is to point your agent at the hosted MCP server. No local server needed.

MCP server: https://www.massblogger.com/api/mcp
Authorization: Bearer <your MCP token>

Generate your token in Massblogger → Settings → Agentic SEO → Hosted MCP access. Each token is scoped to a single user account and its websites.

Claude — CLI (claude.ai)

The easiest way if you use Claude on the command line. Two commands and you're done — Claude will open a browser tab to authenticate automatically.

Step 1 — Remove any old Massblogger connection (safe to skip if first time)

claude mcp remove massblogger

Step 2 — Add the Massblogger MCP server

claude mcp add --transport http massblogger https://www.massblogger.com/api/mcp

Step 3 — Start Claude and verify

claude
# Then ask Claude: "list my websites"

When you run the add command, Claude will open a browser window and ask you to sign in to Massblogger to grant access. No manual token copy-pasting needed. Once authorised, Claude can manage all your WordPress sites from the terminal.

Claude Desktop

Claude Desktop currently works best through mcp-remote. Add this to claude_desktop_config.json:

{
  "mcpServers": {
    "massblogger": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@0.1.38",
        "https://www.massblogger.com/api/mcp",
        "--resource",
        "https://www.massblogger.com/api/mcp",
        "--header",
        "Authorization:${MASSBLOGGER_AUTH_HEADER}"
      ],
      "env": {
        "MASSBLOGGER_AUTH_HEADER": "Bearer <your MCP token>"
      }
    }
  }
}

Use the www host exactly as shown. After saving, restart Claude Desktop and run list_websites as your first tool call to confirm the connection.

Cursor

Use the same mcp-remote config as Claude Desktop in your Cursor MCP settings:

{
  "mcpServers": {
    "massblogger": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@0.1.38",
        "https://www.massblogger.com/api/mcp",
        "--resource",
        "https://www.massblogger.com/api/mcp",
        "--header",
        "Authorization:${MASSBLOGGER_AUTH_HEADER}"
      ],
      "env": {
        "MASSBLOGGER_AUTH_HEADER": "Bearer <your MCP token>"
      }
    }
  }
}

OpenClaw

OpenClaw can connect directly to the hosted MCP server. Replace YOUR_MCP_TOKEN with the token from Massblogger Settings:

openclaw config set mcpServers.massblogger.url "https://www.massblogger.com/api/mcp"
openclaw config set mcpServers.massblogger.headers.Authorization "Bearer YOUR_MCP_TOKEN"
openclaw config set mcpServers.massblogger.name "Massblogger"

If your OpenClaw version uses env vars instead of headers, use the fallback token header:

openclaw config set mcpServers.massblogger.url "https://www.massblogger.com/api/mcp"
openclaw config set mcpServers.massblogger.env.x-massblogger-mcp-token "YOUR_MCP_TOKEN"

After configuration, restart OpenClaw and run list_websites to verify the connection.

Windsurf and other MCP clients

Any client that supports direct HTTP MCP can use the hosted MCP server and pass the token in the Authorization header (or as x-massblogger-mcp-token if the client doesn't support custom headers).

# Windsurf mcp_config.json
{
  "mcpServers": {
    "massblogger": {
      "serverUrl": "https://www.massblogger.com/api/mcp",
      "headers": {
        "Authorization": "Bearer <your MCP token>"
      }
    }
  }
}

# Fallback — some clients only support env vars:
x-massblogger-mcp-token: <your MCP token>

Local / self-hosted MCP server

If you run Massblogger yourself, you can also launch the MCP server locally (stdio mode) and scope it to a single user:

MASSBLOGGER_MCP_USER_EMAIL="you@example.com" npm run mcp:agentic-seo

Then point your MCP client at the local stdio process instead of the HTTP endpoint:

# Claude Desktop / Cursor — local stdio config
{
  "mcpServers": {
    "massblogger-local": {
      "command": "node",
      "args": ["scripts/mcp-agentic-seo.mjs"],
      "cwd": "/path/to/your/massblogger",
      "env": {
        "MASSBLOGGER_MCP_USER_EMAIL": "you@example.com"
      }
    }
  }
}

# OpenClaw — local stdio config
openclaw config set mcpServers.massblogger-local.command "node"
openclaw config set mcpServers.massblogger-local.args '["scripts/mcp-agentic-seo.mjs"]'
openclaw config set mcpServers.massblogger-local.env.MASSBLOGGER_MCP_USER_EMAIL "you@example.com"

Prerequisites & WordPress permissions

• WordPress Application Password: the stored wpAppPassword must belong to a user with the Administrator or Editor role. Author-level credentials can create and edit their own posts but cannot restore trashed content, manage plugins, or modify other users' posts. Generate it in WordPress under Users → Profile → Application Passwords.
• Trash operations (restore_wordpress_post, restore_wordpress_page): require Editor or Administrator role. The WordPress REST API blocks status updates on trashed items for lower-privilege users. If you see an "Unauthorized" error on restore, check that the Application Password belongs to an Editor or Administrator account.
• Plugin management (activate_wordpress_plugin, deactivate_wordpress_plugin, list_wordpress_plugins): requires Administrator role and WordPress 5.5+.
• User management (create_wordpress_user, delete_wordpress_user): requires Administrator role.
• Yoast SEO fields (metaTitle, metaDescription): require the Yoast SEO plugin installed on the WordPress site.
• Menu write operations (create_wordpress_menu, add_wordpress_menu_item, etc.): require a plugin that exposes the WP REST API menus endpoint, such as WP REST API Menus or WordPress 6.x+ Full Site Editing.
• Webhook sites: must have webhook delivery enabled in Massblogger website settings.
• Next.js / REST sites: must be connected in Massblogger so saved posts are accessible via the existing API.
• MCP token: generate in Massblogger Settings → Agentic SEO → Hosted MCP access. No separate API key needed. The token is scoped to your account only.

Partial updates — no full body required

Every update tool is designed for surgical changes. You only send the fields you want to modify:

// Change only the title and status of post 42
{
  "name": "update_post",
  "arguments": {
    "website": "myblog.com",
    "wpPostId": 42,
    "title": "10 Best Lightweight Camping Stoves in 2026",
    "status": "publish"
  }
}

// Change only the Yoast meta description
{
  "name": "update_post",
  "arguments": {
    "website": "myblog.com",
    "wpPostId": 42,
    "metaDescription": "We tested 10 stoves so you don't have to. Here's what we found."
  }
}

Fields you don't include are left unchanged. Use get_wordpress_post first if the agent needs to read the existing content before making changes.

Full tool catalog

Recommended agent workflow

• Start with list_websites to discover available sites, check which ones are WordPress-connected, and get the correct website ID or domain.
• Use get_site_context so the agent reads the site's llms.txt / audience and writing rules before generating content.
• Use list_wordpress_categories and list_wordpress_tags to resolve names to IDs before assigning them to posts.
• For content updates, call get_wordpress_post first to read the current title and content, then call update_post with only the changed fields.
• To add an image inside an article, use insert_image_into_post — no need to rewrite the whole body. Use search_replace_post_content for targeted text changes.
• For featured images: use generate_and_set_featured_image to generate an AI image and set it in one call, or list_massblogger_media + use_massblogger_media_in_wordpress to reuse an image already in the Massblogger library.
• Use set_wordpress_featured_image when you already have a WordPress media ID and just need to assign it.
• Use upload_wordpress_media to push images from any public URL and get back a media ID.
• To restore accidentally trashed content, use restore_wordpress_post or restore_wordpress_page. To undo a content edit, use list_post_revisions then restore_post_revision — WordPress saves a revision on every save, so you can always roll back.
• For production safety, run set_publish_guard on any site where you don't want the agent publishing directly. All status: publish calls will be silently redirected to draft at the server level — the agent cannot override it.
• For portfolio-wide changes, use search_replace_across_sites with dryRun: true first, then re-run without it once you're happy.
• Use update_wordpress_seo after publishing to set Yoast SEO title and meta description.
• Use get_activity_logs at the end of any session to audit what the AI actually did.

Example agent prompts

Research topics for myblog.com about "best lightweight camping stoves".
Save the best 3 ideas. Generate a full article from the top topic and publish it.
Then generate and set a featured image for that post.
Set the Yoast meta description, then show me the activity log.

List the last 20 published posts on myblog.com.
For each one that has no featured image, generate an AI featured image using the post title as context and set it.
Then insert a relevant image after the first paragraph of each post.

List the last 20 published posts on myblog.com.
For each one that's missing a Yoast meta description, write a compelling
150-character description and update it using update_wordpress_seo.

Search all posts on myblog.com that mention "BestGadgets 2024" and replace with "BestGadgets 2026".
Use search_replace_post_content for each post.

List all categories on myblog.com.
Create a new parent category called "Gear Reviews".
Move "Tents", "Stoves", and "Sleeping Bags" to be children of "Gear Reviews".
Confirm the updated category tree.

We rebranded from "GearLife" to "TrailPro". 
Run search_replace_across_sites with dryRun: true first to show me how many posts and pages 
would change across all my sites, then run it for real once I confirm.

Create a WordPress user named "Sarah Chen" (sarah@example.com) 
with the role "author" on these 5 sites: myblog.com, gearblog.com, hikingblog.com, campblog.com, trailblog.com.

I just ran a search/replace on post 142 on myblog.com and it broke something.
List the last 5 revisions for that post, then restore it to the revision just before the most recent one.

Before running the content refresh job, enable publish guard on myblog.com.
Then generate and update the 10 oldest posts with new content and updated Yoast meta.
Show me the drafts when done so I can review before publishing.

WordPress vs webhook vs Next.js

Related guides

• WordPress Application Passwords — how to generate and store credentials
• Webhooks Guide — custom site delivery and signature verification
• pSEO Webhooks — programmatic SEO page delivery
• REST API — frontend rendering for API-backed sites