MCP Server for CAP
Enables Claude to interact with a document processing backend, providing tools for listing, retrieving, uploading, and processing documents, as well as resources for pending documents and statistics.
README
Build a Production-Grade MCP Server for CAP
Overview
The server exposes a set of MCP tools and resources that let Claude interact with a document processing backend. It uses the MCP SDK with a StreamableHTTPServerTransport, meaning each session is stateful and tracked by a mcp-session-id header.
Authentication
The token is never exposed to the client — it lives only in the server process via process.env.BEARER_TOKEN. The MCP endpoint itself (POST /mcp) does not require authentication from Claude's side
Tool Reference
list_documents : Lists all documents, with an optional status filter.
get_document : Retrieves full details for a single document, including its parsed invoice and line items.
upload_document : Uploads a document as a base64-encoded string to the backend.
process_document : process_document
get_audit_log : Returns all audit log entries for a given document.
Resources
documents://pending : List of documents with PENDING status
documents://stats : Returns aggregate statistics across all documents:
Total document count
Count of DONE documents
Count of FAILED documents
Average AI confidence score (from expanded invoice data)
Retry Strategy
All OData calls go through withRetry(), an exponential backoff wrapper:
Attempt 1 ──► fail ──► wait 500ms
Attempt 2 ──► fail ──► wait 1000ms
Attempt 3 ──► fail ──► throw
Connecting from SAP BTP / Cloud Foundry
Create an Node.js application. npm install .env.example .env # fill in APP_URL and BEARER_TOKEN node index.js # To run the application locally
Push to CF
cf push # To push the code to CF After deploying the application i tested from the POSTMAN.
NOTE : By using the VScode, I cloned the application and i tested via the official MCP Inspector. node index.js # To run the application locally. For MCP Inspector # npx @modelcontextprotocol/inspector.
NOTE : When iam trying to connect the Claude code with my application, It is not working. (It may work with a licensed Claude account).
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。