WP-MCP
Enables interaction with WordPress sites through the WordPress REST API, dynamically exposing all routes as MCP tools for content management and site configuration.
README
WP-MCP
MCP server that wraps the WordPress REST JSON API. It dynamically discovers all available routes from your WordPress site and exposes each as an MCP tool.
Features
- Automatic route discovery from
/wp-json— all core and plugin endpoints are exposed - Authentication via WordPress Application Passwords
- Optional SSL certificate verification bypass for local/dev environments
- One tool per route with
methodas a parameter (GET, POST, PUT, PATCH, DELETE) - Clean tool names:
posts,posts.id,categories,media.id, etc.
Configuration
The server is configured via environment variables passed through your MCP client:
| Variable | Required | Description |
|---|---|---|
WP_URL |
Yes | WordPress site URL (e.g. https://example.com) |
WP_USERNAME |
No | WordPress username for authenticated requests |
WP_APP_PASSWORD |
No | WordPress Application Password |
WP_IGNORE_SSL |
No | Set to "true" to skip SSL certificate verification |
WP_TOOL_MODE |
No | Tool exposure strategy: all (default), compact, allowlist, or blocklist |
WP_TOOL_FILTER |
No | Comma-separated list of tool names or route patterns for allowlist/blocklist modes |
WP_DESCRIPTION_MODE |
No | verbose or minimal (default). Minimal strips property descriptions to reduce token usage |
Generating an Application Password
- In your WordPress admin, go to Users > Profile
- Scroll to Application Passwords
- Enter a name (e.g. "wp-mcp") and click Add New Application Password
- Copy the generated password
Setup
Build from source
npm install
npm run build
MCP client configuration
Add to your MCP client settings (e.g. Claude Desktop claude_desktop_config.json or VS Code settings.json):
{
"mcpServers": {
"wordpress": {
"command": "node",
"args": ["/path/to/wp-mcp/dist/index.js"],
"env": {
"WP_URL": "https://example.com",
"WP_USERNAME": "admin",
"WP_APP_PASSWORD": "xxxx xxxx xxxx xxxx",
"WP_IGNORE_SSL": "false"
}
}
}
}
How it works
On startup the server fetches the WordPress REST API index at WP_URL/wp-json, which returns all registered routes with their methods, arguments, and schemas. Each route becomes an MCP tool:
| WP Route | Tool Name |
|---|---|
/wp/v2/posts |
posts |
/wp/v2/posts/(?P<id>[\d]+) |
posts.id |
/wp/v2/categories |
categories |
/wp/v2/media/(?P<id>[\d]+) |
media.id |
/wc/v3/products |
wc.v3.products |
The wp/v2/ prefix is stripped from core routes for cleaner names. Plugin namespaces are preserved to avoid collisions.
Every tool accepts a required method parameter (enum of the HTTP methods that route supports) plus the route's own arguments as additional parameters.
Built-in parameters
These parameters are injected into every tool's schema by the MCP server.
_fields — Added to all tools. Comma-separated list of fields to include in the response (e.g. id,title,slug). This is a native WordPress query parameter that reduces response size — it is passed through to WordPress.
The following parameters are intercepted by the MCP server and not sent to WordPress:
_save_response — Added to all tools. File path to save the response to instead of returning it inline. When set, the tool returns a compact summary (file path, size, and identifying fields like id/title/type) instead of the full response body. Useful for large responses that would waste context tokens.
_save_response_field — Added to all tools. Used with _save_response. Dot-notation path to extract a specific field before saving (e.g. content.rendered). Only the extracted value is written to the file — string values are written raw (not JSON-quoted), so you get clean HTML/text.
_file_params — Added to writable tools (POST/PUT/PATCH/DELETE). An object mapping parameter names to file paths. Each file is read and its contents used as the string value for that parameter. Example: {"content": "/tmp/page.html", "excerpt": "/tmp/excerpt.txt"}.
_body_file — Added to writable tools. Path to a JSON file whose contents are parsed and merged into the request body. Precedence: _file_params > explicit inline parameters > _body_file.
File-based content workflow
When working with large content (e.g. editing a WordPress page), agents can use these parameters to keep content out of the conversation context entirely:
1. GET pages.id { "id": 42, "_save_response": "/tmp/page.json" }
→ returns: {"saved_to":"/tmp/page.json","size":"24.3KB","id":42,"title":"About Us","type":"page"}
2. Agent reads /tmp/page.json, edits it locally
3. POST pages.id { "id": 42, "method": "POST", "_body_file": "/tmp/page.json" }
→ MCP reads the file and sends its contents as the request body
Or to work with just the content field:
1. GET pages.id { "id": 42, "_save_response": "/tmp/content.html", "_save_response_field": "content.rendered" }
→ saves raw HTML to file, returns compact summary
2. Agent edits /tmp/content.html
3. POST pages.id { "id": 42, "method": "POST", "_file_params": { "content": "/tmp/content.html" } }
→ MCP reads the HTML file and sends it as the content parameter
Other features
- Media uploads: The
mediatool accepts afile_pathparameter for uploading local files. - ACF support: When Advanced Custom Fields is detected, writable tools on core routes get an
acfobject parameter for setting custom field values. refresh_tools: A built-in tool that re-discovers all WordPress REST API routes. Use after installing plugins or registering new post types.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。