mcp-easycompliance

mcp-easycompliance

Enables AI agents to screen business partners against EU/US/HADDEX sanctions and PEP lists, and maintain the easycompliance daily monitoring list.

Category
访问服务器

README

mcp-easycompliance

An MCP (Model Context Protocol) server for easycompliance – the German SaaS platform for automated sanctions list and PEP screening.

It lets AI agents (Claude Desktop, Claude Code, Cursor, agent frameworks and other MCP clients) screen business partners against EU/US/HADDEX sanctions lists and PEP lists, and maintain the daily monitoring list of an easycompliance account. The server is a pure consumer of the easycompliance REST API – executed tool calls (screenings, status/hit queries and executed monitoring changes) count towards the account's regular API quota and appear in the account's API log. Rejected requests (e.g. an invalid key) and deletion attempts without a matching entry are not logged.

Tools

Tool Kind Description
check_sanctions_list(name, accuracy?) read-only One-time screening of a name against the EU/US/HADDEX sanctions lists
check_pep_list(name, accuracy?) read-only One-time screening of a name against the PEP list (politically exposed persons)
get_hits_last_24h(list) read-only Hits produced by the daily monitoring within the last 24 hours
get_list_status(list) read-only Date (dd.mm.yyyy) of the last update of the selected list
check_and_add_to_monitoring(list, name, ref?, duplicate_by?, accuracy?) write Screens the name and adds it to the account's daily monitoring list
remove_from_monitoring(list, by, value) write Removes one matching entry from the daily monitoring list (by name or by reference)

list is "sanctions" or "pep". Screening results are always structured: hit (boolean), hitCount and hits[] with a percent match score – an empty result (HTTP 204 of the API) is returned as hit: false, never as an error.

Compliance notes

  • The tool results are the authoritative source. The tool descriptions instruct the calling model to never soften, filter or omit hits.
  • If hit is true, the professional assessment of the match belongs in the easycompliance customer portal (https://kunde.easycompliance.de, Monitoring → Hits) – not in the AI conversation.
  • check_and_add_to_monitoring has a persistent side effect (the name is re-screened daily and counts towards the account's monitoring quota); remove_from_monitoring removes monitoring entries. Both are clearly marked as write tools (no readOnlyHint), so MCP clients can require confirmation.
  • remove_from_monitoring removes exactly one matching entry per call and is deliberately not marked idempotent: if several monitoring entries share the same name, a retried call removes another one of them. Use by="ref" with unique references to address a specific entry.

Requirements

You need an easycompliance account with API access. Customer service (https://www.easycompliance.de/) provides:

  • your API key,
  • the sanctions list API URL and the PEP API URL (only needed if your account uses customer-specific URLs – the public defaults are built in).

Configuration

All configuration is via environment variables. The MCP server itself never writes the key to a file and never logs it. Note that your MCP client stores the values you configure (including EC_API_KEY, as in the examples below) in its own configuration file, e.g. claude_desktop_config.json – protect that file like a password, and prefer your client's environment-variable or secret-management support where available.

Variable Required Default Description
EC_API_KEY yes (stdio) API key provided by easycompliance customer service
EC_SANCTIONS_URL no https://www.easycompliance.de/easy.api Sanctions list API endpoint
EC_PEP_URL no https://www.easycompliance.de/pep.api PEP API endpoint

Installation

The package runs with Node.js >= 20 and is started via npx mcp-easycompliance (stdio transport).

Claude Desktop

Add the server to claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "easycompliance": {
      "command": "npx",
      "args": ["-y", "mcp-easycompliance"],
      "env": {
        "EC_API_KEY": "your-api-key"
      }
    }
  }
}

Claude Code

claude mcp add easycompliance --env EC_API_KEY=your-api-key -- npx -y mcp-easycompliance

Cursor

Add to ~/.cursor/mcp.json (or the project's .cursor/mcp.json):

{
  "mcpServers": {
    "easycompliance": {
      "command": "npx",
      "args": ["-y", "mcp-easycompliance"],
      "env": {
        "EC_API_KEY": "your-api-key"
      }
    }
  }
}

Other MCP clients

Any MCP client that supports the stdio transport can run npx -y mcp-easycompliance with the EC_API_KEY environment variable set.

Remote endpoint (streamable HTTP)

easycompliance hosts a remote endpoint with the identical tool set at https://www.easycompliance.de/mcp.api (streamable HTTP; implemented in PHP inside the easycompliance stack, verified against the same MCP SDK client test harness as this package). It works with remote clients that can send custom request headers, such as the OpenAI Responses API or MCP SDK clients; ChatGPT custom connectors depend on compatible authentication (OAuth is a planned extension, see below), so ChatGPT app support is in preparation.

The package also contains a second entry point, mcp-easycompliance-http, which serves the same tool set over streamable HTTP for self-hosting (e.g. enterprise deployments in your own infrastructure – see deploy/):

  • Authentication: the client sends its easycompliance API key in the X-API-Key request header; the server passes it through 1:1 to the REST API on every call. There is no key store, no session state and no persistence in the MCP layer.
  • OAuth is currently out of scope (planned as a future extension for connector directories that require it).

Example (Claude Code):

claude mcp add --transport http easycompliance https://www.easycompliance.de/mcp.api --header "X-API-Key: your-api-key"

Operations & security (self-hosting the HTTP endpoint)

Deployment templates (nginx reverse proxy, systemd unit, Dockerfile + step-by-step guide) are in deploy/:

  • The Node process binds to 127.0.0.1 only; TLS and the public hostname are provided by the nginx reverse proxy.
  • The server is stateless: each request creates a fresh server instance with the key from the request header; nothing outlives the request.
  • The MCP layer logs no names, no screening data and no keys – the API's own log (apilog) is the audit trail for executed calls, which also count towards the account's API quota there.
  • GET /healthz returns 200 ok for uptime checks.

Environment variables of the HTTP entry point: MCP_HTTP_HOST (default 127.0.0.1), MCP_HTTP_PORT (default 8765), MCP_HTTP_PATH (default /mcp), plus EC_SANCTIONS_URL/EC_PEP_URL as above. EC_API_KEY is not used in HTTP mode – keys always come from the client request.

API behaviour

The underlying REST API responds with HTTP 200 (JSON result array), 204 (no hit / no entry), 400 (invalid request body), 401 (invalid API key or suspended account), 403 (operation not included in the API plan) or 405 (non-POST). The server normalizes these into structured tool results and clear tool errors; an HTTP 200 without a result array is reported as a transport error and never as a clean "no hit". For get_list_status, an HTTP 204 (no completed list update verifiable yet) and a malformed date are reported as tool errors rather than as an empty status.

Known limitation of the API contract (shared with all API clients, including the n8n community node): for check_and_add_to_monitoring the API reports the screening result without a separate confirmation of the monitoring insertion – a server-side insert failure is logged on the server but not signalled in the HTTP response. Extending the API contract is outside the scope of this consumer package.

Development

npm ci
npm run lint
npm run build
EC_API_KEY=<test-key> npm run livetest        # live harness (stdio logic) against the production API
EC_API_KEY=<test-key> npm run livetest:http   # live harness for the streamable HTTP transport

Resources

License

MIT

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选