mcp-invoices
An MCP server for managing invoices with DynamoDB, offering tools to create, retrieve, and list invoices, plus a REST API.
README
mcp-hello
Hello world MCP server, structured just like mcp: McpServer from the official SDK, stdio/HTTP transports, and AWS Lambda deploy via the Serverless Framework. Also includes an invoices ("facturas") feature backed by DynamoDB, exposed both as MCP tools and as a REST API.
Install
npm install
npm run build
Local (stdio)
npm run mcp— starts the server over stdio (tsx src/server.ts)
To connect it to Claude Code / Cursor, point the MCP client at the npx tsx src/server.ts command (or the compiled binary in dist/server.js after npm run build).
Local (HTTP)
npm run mcp:http
- MCP:
http://localhost:8787/mcp - Health:
http://localhost:8787/health - Invoices REST API:
http://localhost:8787/invoices
Requires AWS credentials with access to the InvoicesTable DynamoDB table (see Invoices below). Set INVOICES_TABLE_NAME (and optionally AWS_REGION / DYNAMODB_ENDPOINT for DynamoDB Local) in your .env.
Deploy (AWS Lambda)
npm run deploy -- --stage dev
npm run deploy:remove -- --stage dev
The deploy uses serverless-esbuild to bundle lambdas/handler.ts and exposes the MCP server through API Gateway (HTTP API). Once it finishes, serverless deploy prints the public URL (https://<id>.execute-api.<region>.amazonaws.com/mcp); add serverless-domain-manager and your own DNS config if you want a custom domain.
Scripts
| Script | Description |
|---|---|
npm run mcp |
MCP over stdio (local) |
npm run mcp:http |
MCP over HTTP (local) |
npm run build |
Compiles src/ to dist/ |
npm run ts:check |
Type-checks without emitting |
npm run deploy |
Lambda + API Gateway |
npm run deploy:remove |
Removes the deployed stack |
Tools
say_hello({ name? })→ returnsHello, <name>!(orHello, world!if no name is given).save_invoice({ invoiceNumber?, date, provider, concept, currency, amount, status?, items? })→ creates an invoice in DynamoDB.get_invoice({ id })→ fetches a single invoice by id.list_invoices({ from?, to?, limit?, cursor? })→ lists invoices sorted by date (newest first), optionally filtered by date range (YYYY-MM-DD).
Invoices
Invoices are stored in a DynamoDB table (InvoicesTable, provisioned in serverless.yml) with id as the primary key and a byDate global secondary index for date-range queries.
Fields: id, invoiceNumber?, date (ISO YYYY-MM-DD), provider, concept, currency (default USD), amount, status (pending | paid, default pending), items? ({ description, quantity, unitPrice }[]), createdAt, updatedAt.
REST API
| Method & path | Description |
|---|---|
POST /invoices |
Create an invoice. Body matches the invoice fields above (minus id/createdAt/updatedAt). |
GET /invoices |
List invoices. Query params: from, to (ISO dates), limit (default 20, max 100), cursor (pagination token from a previous response). |
GET /invoices/:id |
Fetch a single invoice, 404 if not found. |
No authentication yet — add it before exposing this outside a trusted network.
Architecture
MCP client → src/server.ts (stdio) | src/index.ts / lambdas/handler.ts (HTTP) → src/register-server.ts → tools/
REST client → src/app.ts (Express) → src/routes/ → src/repositories/ → DynamoDB
| File | Role |
|---|---|
src/register-server.ts |
Factory that creates the McpServer and registers the tools |
src/server.ts |
Stdio entrypoint (Cursor/Claude Code, local) |
src/app.ts |
Express app with the Streamable HTTP transport (/mcp, /health, /invoices) |
src/index.ts |
Local HTTP entrypoint (npm run mcp:http) |
lambdas/handler.ts |
@codegenie/serverless-express wrapper for AWS Lambda |
src/tools/ |
One register*Tool function per MCP tool |
src/routes/ |
Express routers for the REST API |
src/repositories/ |
DynamoDB data access (shared by MCP tools and REST routes) |
src/models/ |
Zod schemas / types for domain entities |
src/db/dynamo-client.ts |
Shared DynamoDBDocumentClient + table/index names |
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。