WebsiteOps MCP Server
Bridges AI coding assistants with website codebases, local git repositories, and shared hosting via FTP/SFTP, enabling code discovery, file management, git operations, smart deployments, backups, and rollbacks.
README
WebsiteOps MCP Server
WebsiteOps MCP is a production-grade, local-first Model Context Protocol (MCP) server designed to bridge AI coding assistants (ChatGPT, Claude Desktop, Cursor, VS Code, Windsurf, Gemini CLI, etc.) with website codebases, local git repositories, and shared hosting / cPanel environments via FTP/SFTP.
With WebsiteOps, AI assistants can discover structure, analyze functions/classes/routes/schemas, monitor file changes, run git commits/pulls/pushes, perform smart diff-based deployments, create backups, and rollback deployments—all locally and offline.
Key Features
- Website Discovery: Analyzes codebases to identify language, framework (Laravel, CodeIgniter, Express, React, Next.js), database type, and stats (route/form/API counts).
- Path Security & Isolation: Strict validation to ensure no read/write operations escape the configured site root.
- Encrypted Credentials: AES-256-GCM secure storage for FTP/SFTP credentials.
- Offline Semantic Search: Built-in TF-IDF and structure matching index for offline concepts search (e.g. "Find payment gateway", "Find registration flow").
- Smart Deployments:
- Detects only modified/added files.
- Automatically downloads current remote versions to local backup folders before pushing changes.
- Synchronizes deletes if desired.
- Supports 1-click rollback restoring the remote server state from the backup folder.
- Background File Watching: Chokidar-driven file watching automatically rebuilds the code index when files change.
- Git wrapper: Runs standard git commands directly through the AI.
Installation & Setup
1. Prerequisites
- Node.js (v18 or higher recommended)
- Git installed and available in system PATH
2. Setup Server
Clone or copy this project to your directory (e.g. C:\mcp-servers\websiteops-mcp), then run:
npm install
npm run build
This compiles TypeScript code from src/ to dist/.
Multi-Website Configuration
You can manage unlimited websites. Website configurations are stored locally in SQLite (.websiteops/metadata.db) with sensitive passwords and private keys encrypted.
Registering a Site
To add a site, call the add_site tool. Note that only name and localPath are required. All FTP settings are optional:
{
"name": "project1",
"localPath": "/Users/username/Projects/my-site"
}
If you also wish to configure FTP/SFTP deployment for the site:
{
"name": "project1",
"localPath": "/Users/username/Projects/my-site",
"remotePath": "/public_html",
"ftpHost": "ftp.example.com",
"ftpPort": 21,
"ftpUser": "ftp_username",
"ftpPassword": "secure_ftp_password",
"ftpType": "ftp"
}
For SFTP with Private Key:
{
"name": "project2",
"localPath": "C:/Projects/my-next-site",
"remotePath": "/var/www/next",
"ftpHost": "ssh.example.com",
"ftpPort": 22,
"ftpUser": "ubuntu",
"ftpType": "sftp",
"privateKey": "-----BEGIN OPENSSH PRIVATE KEY-----\n..."
}
Client Configuration Examples
Add the server to your preferred MCP host configuration:
Claude Desktop
Add this to your claude_desktop_config.json (located at %APPDATA%\Claude\claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"websiteops": {
"command": "npx",
"args": [
"-y",
"websiteops-mcp"
],
"env": {
"WEBSITEOPS_SECRET_KEY": "YOUR_SUPER_SECRET_KEY"
}
}
}
}
Cursor, VS Code, Windsurf
Add a new MCP server in the settings panel:
- Type:
stdio - Command:
npx -y websiteops-mcp - Environment variables:
WEBSITEOPS_SECRET_KEY=YOUR_SUPER_SECRET_KEY
Alternatively, you can run a locally cloned server:
- Command:
node /path/to/cloned/websiteops-mcp/dist/index.js
SQLite Metadata Schema Details
WebsiteOps stores data locally inside a .websiteops folder created in the server's working directory:
- sites: Stores paths, connection details, and credentials (encrypted).
- deployments: Tracks deployment status, timestamps, commit hashes, errors, and local backup folder directories.
- code_index: Extracted classes, functions, routes, forms, payments, and authentication structures.
- search_index: Document terms (TF-IDF tokens) for offline text queries.
MCP Tools Reference
| Tool Name | Parameters | Description |
|---|---|---|
add_site |
name, localPath, remotePath?, ftpHost?, ftpPort?, ftpUser?, ftpPassword?, ftpType?, privateKey? | Register a new website profile (FTP configuration is optional) |
list_sites |
List all configured website profiles | |
remove_site |
id | Remove a site connection profile |
discover_website |
siteName, siteId, localPath | Auto-detect framework, language, DB, count of routes |
get_project_tree |
siteName, siteId, maxDepth | Return file/folder listing of project |
read_file |
siteName, filePath | Read contents of a local file safely |
write_file / create_file |
siteName, filePath, content | Edit or create local files safely |
delete_file |
siteName, filePath | Delete a local file |
search_files / search_code |
siteName, query | Find file name patterns or search text inside files |
search_functions / search_classes / search_routes |
siteName, query | Find classes, functions or routes indexed in codebase |
git_status / git_diff / git_commit |
siteName, options/message/files | Core Git actions wrapper |
git_pull / git_push / git_sync |
siteName | Pull/Push/Sync commits with remote repository |
ftp_connect / ftp_compare |
siteName | Verify connection or check local vs remote file diffs |
ftp_sync |
siteName | Full directory synchronisation to remote |
deploy_project / deploy_changed_files |
siteName, files, commitHash | Safe deployment with auto-backup of remote files before upload |
rollback_deployment |
deploymentId | Restore the remote file state from a previous backup |
deployment_history |
siteName | View previous deployment attempts and logs |
semantic_search |
query, siteName | Local offline search for terms (e.g. login, Payment gateways) |
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。