mcp-for-apache-ofbiz
A proof of concept MCP server for Apache OFBiz
README
MCP Server for REST APIs
This project provides a prototype implementation of an MCP server that:
- exposes specific tools,
- receives requests from an MCP client (usually hosted in a generative AI application such as Claude Desktop),
- forwards those requests to a remote backend via RESTful API endpoints,
- is implemented using the Anthropic SDK for TypeScript.
The server enables generative AI applications to interact with remote systems that expose REST API endpoints, such as Apache OFBiz and Moqui.
The server is implemented in two versions, one that runs as a local MCP server (stdio transport) and one that runs as a remote MCP server (Streamable HTTP transport).
The project is implemented in TypeScript, uses the Anthropic TypeScript SDK, and requires:
- Node.js
- npm
Table of Contents
- Features
- Configuration
- Project Structure
- Build the Project
- Test the Local MCP Server
- Test the Remote MCP Server
- Inspect the MCP servers
Features
The project includes two alternative MCP servers:
- Local MCP server (
src/server.ts) — communicates with the MCP client via stdio transport. - Remote MCP server (
src/server-remote.ts) — communicates with the MCP client via MCP Streamable HTTP transport.
The servers are modular and dynamically discover MCP tools contained in the tools directory.
Each tool is defined and implemented in its own file. For example, the sample tool tools/findProductById.ts invokes an endpoint in Apache OFBiz to retrieve product information for a given ID. This works with an out-of-the-box (OOTB) OFBiz instance with the rest-api plugin installed.
Configuration
Server configuration is managed via config/config.json, which defines:
BACKEND_API_BASE— the base URL for backend REST API callsBACKEND_API_AUTH- the URL to get the OFBiz APIs access tokenBACKEND_AUTH_TOKEN— the token used to authorize backend API callsSERVER_PORT— the port on which the MCP server listens for client connections (required only for the remote server)
The authorization token for the OFBiz API can be easily generated and set up by running the script:
update_token.sh <user> <password>
This script retrieves a JWT for an OOTB OFBiz instance (e.g., https://demo-stable.ofbiz.apache.org/rest/auth/token).
Project Structure
mcp-prototypes/
├── config/
│ └── config.json # Server configuration (backend API base, auth token, etc.)
├── src/
│ ├── server.ts # Local MCP server (stdio transport)
│ ├── server-remote.ts # Remote MCP server (Streamable HTTP transport)
│ ├── toolLoader.ts # Loader of tool definitions from "tools/"
│ └── tools/
│ └── findProductById.ts # Example tool calling an Apache OFBiz REST endpoint
├── update_token.sh # Script to refresh backend auth token
├── package.json
├── tsconfig.json
└── README.md
Build the Project
npm install
npm run build
Test the Local MCP Server
You can test the local MCP server with Claude Desktop.
Edit or create the Claude Desktop configuration file:
~/Library/Application\ Support/Claude/claude_desktop_config.json
Add your local MCP server configuration:
{
"mcpServers": {
"Apache OFBiz": {
"command": "node",
"args": ["PATH_TO/mcp-prototypes/build/server.js"]
}
}
}
After updating the configuration file, launch Claude Desktop and try the following sample prompts:
- "Can you provide some information about the product WG-1111?"
- "Can you provide some information about a product?"
(Claude will ask for a product ID before invoking the tool.) - "Can you compare two products?"
(Claude will ask for two product IDs, invoke the tool twice, and then compare the results.)
Test the Remote MCP Server
Start the server:
node build/server-remote.js
You can test the local MCP server with Claude Desktop.
Edit or create the Claude Desktop configuration file:
~/Library/Application\ Support/Claude/claude_desktop_config.json
Add your local MCP server configuration:
{
"mcpServers": {
"Apache OFBiz": {
"command": "npx",
"args": ["-y", "mcp-remote", "http://localhost:3000/mcp", "--allow-http"]
}
}
}
Inspect the MCP servers
You can use Anthropic’s Inspector to easily test interactions with the local and remote MCP servers. You can do this also when a remote server is executed in your local host or private network, without requiring valid certificates or deploying the server on a publicly accessible host.
Run (and install) the Inspector with:
npx @modelcontextprotocol/inspector
This will open a browser window ready to test your MCP servers.
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。