cebelca-mcp
An MCP server that wraps the cebelca.biz accounting API, exposing tools for operations like managing partners, invoices, proformas, and fetching PDFs.
README
cebelca-mcp
An MCP server that wraps the cebelca.biz accounting API, exposing it as tools an MCP client (Claude, etc.) can call.
Development
This project uses devenv for its development
environment (Node.js, yarn). With direnv installed, cd into the directory and
the environment loads automatically; otherwise run devenv shell.
The cebelca.biz API key is read from the CEBELCA_BIZ_API_KEY environment
variable. It is set in .envrc (git-ignored) and loaded by direnv/devenv. The
server refuses to start if the variable is missing.
Optional: CEBELCA_BIZ_BASE_URL overrides the API base URL (defaults to
https://www.cebelca.biz/API).
Commands
yarn install # install dependencies
yarn dev # run from source with tsx (no build step)
yarn build # compile TypeScript to dist/
yarn start # run the compiled server (dist/index.js)
yarn typecheck # type-check without emitting
Project layout
src/
index.ts Entry point: builds the server, connects over stdio.
config.ts Loads config from the environment (API key, base URL).
client.ts HTTP client for the cebelca.biz resource/method API.
tools.ts MCP tool registration. Add new tools here.
The cebelca.biz API
The API is a resource/method RPC over HTTP (docs: Workonomic-API-bash, making-an-invoice):
- Every call is an HTTP
POSTwithx-www-form-urlencodedarguments. - Resource and method go in the
_rand_mquery params (_m2stacks a second method,_fpicks the output format,_x=1is explore mode). - Auth is HTTP Basic: the API token is the username, the password is the
literal
x. - Success responses look like
["ok", [{...}]]or[[{...}]]; errors like["validation", {...}]or[[{"err": "..."}]].CebelcaClientunwraps the success payload and throwsCebelcaErroron failures. - PDFs come from a separate
.../API-pdfendpoint.
Tools
| Tool | Purpose |
|---|---|
cebelca_explore |
Discover resources / methods / arguments (explore mode). |
cebelca_call |
Generic escape hatch: call any resource/method directly. |
cebelca_assure_partner |
Find-or-create a partner, return its id. |
cebelca_create_invoice_head |
Create an issued-invoice head (insert-smart-2). |
cebelca_add_invoice_line |
Add a body line to an invoice. |
cebelca_mark_invoice_paid |
Mark an invoice paid (by id or external id). |
cebelca_finalize_invoice |
Issue an invoice — noncash or fiscal (FURS). |
cebelca_get_fiscal_info |
Fetch ZOI/EOR/QR fiscal details. |
cebelca_create_proforma_head |
Create a proforma (pre-invoice) head. |
cebelca_add_proforma_line |
Add a body line to a proforma. |
cebelca_proforma_to_invoice |
Convert a proforma into an invoice. |
cebelca_get_pdf |
Fetch a document PDF as a base64 resource. |
Adding tools
Register new tools in src/tools.ts via server.registerTool(...), using
CebelcaClient.call(resource, method, args) and the run helper to turn
responses (and CebelcaErrors) into MCP tool results. When you need an
operation no dedicated tool covers, cebelca_explore + cebelca_call reach the
whole API.
Using with an MCP client
Build first (yarn build), then point your client at the compiled entry:
{
"mcpServers": {
"cebelca": {
"command": "node",
"args": ["/absolute/path/to/cebelca-mcp/dist/index.js"],
"env": { "CEBELCA_BIZ_API_KEY": "your-key-here" }
}
}
}
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。