FattureInCloud MCP
MCP server for the Fatture in Cloud API v2 that lets Claude read and manage issued and received documents (invoices, credit notes, quotes, expenses, etc.) via CRUD operations.
README
FattureInCloud MCP
MCP (Model Context Protocol) server for the Fatture in Cloud API v2. Lets Claude (Code / Desktop / any MCP client) read and manage (CRUD) issued documents (invoices, credit notes, quotes...) and received documents (expenses, passive credit notes...).
Two transports over the same tools: stdio as a local child process, and Streamable HTTP for a shared deployment that keeps no credentials of its own.
Tools
| Tool | Description |
|---|---|
list_companies |
Companies accessible with the token (source of company_id) |
get_company_info |
Config lookups: VAT types, payment methods/accounts, expense categories |
list_issued_documents |
Paginated list of issued documents, one type per call (default invoice) |
get_issued_document |
Full detail of a single issued document |
list_received_documents |
Paginated list of received documents (default expense) |
get_received_document |
Full detail of a single received document |
create_issued_document |
Create an invoice, credit note, quote... |
update_issued_document |
Update an issued document (partial: only the fields passed) |
delete_issued_document |
Delete an issued document (irreversible) |
create_received_document |
Create an expense or other received document |
update_received_document |
Update a received document (partial) |
delete_received_document |
Delete a received document (irreversible) |
upload_attachment |
Upload a local file, returns an attachment_token to bind via create/update |
delete_document_attachment |
Remove a document's attachment (FIC API currently 500s here, see docs) |
See docs/api.md for parameters, payload shapes and API quirks. List tools support:
q— Fatture in Cloud filter language, e.g.date >= '2026-01-01' and date <= '2026-06-30',entity.name like '%acme%',amount_gross > 1000sort— e.g.-date(descending)page/per_page(min 5, max 100) — responses includecurrent_page,last_page,totalfieldset—basic(compact, default for lists) ordetailed(includes line items)
Setup
npm install
npm run build
Authentication
Two values, wherever they come from:
FIC_ACCESS_TOKEN(required) — a manual access token: generate it from the Fatture in Cloud developer area selecting the read scopes for issued and received documents. Manual tokens never expire (revocable from the same page).FIC_COMPANY_ID(optional) — default company ID; if unset, tools require an explicitcompany_idargument (discover it withlist_companies).
Over stdio they are environment variables of the server process. Over HTTP they are the X-FIC-Token and X-FIC-Company headers of each request, so the deployed instance stores neither.
Configuration templates for Claude Code, Claude Desktop and opencode, local stdio and remote HTTP, are in examples/.
Claude Code
A project-scoped .mcp.json (gitignored) holds the credentials. From any other directory:
claude mcp add --scope user fattureincloud -e FIC_ACCESS_TOKEN=<token> -e FIC_COMPANY_ID=<id> -- node <repo>\dist\index.js
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"fattureincloud": {
"command": "node",
"args": ["<repo>\\dist\\index.js"],
"env": {
"FIC_ACCESS_TOKEN": "<token>",
"FIC_COMPANY_ID": "<id>"
}
}
}
}
Remote (Streamable HTTP)
dist/http.js serves the same tools over HTTP and takes the credentials from the X-FIC-Token and X-FIC-Company request headers instead of the environment, so the host running it stores no secret. ./scripts/deploy.ps1 installs it as a systemd service on a remote machine — see docs/deploy.md.
Extending
The API surface is small on purpose. To add endpoints (clients, suppliers, products, receipts, taxes, cashbook...), follow the pattern in src/server.ts: one registerTool per action calling ficGet from src/fic.ts. The full OpenAPI spec lives at fattureincloud/openapi-fattureincloud.
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。