Turning four compliance and lead-cost checks into a local, read-only MCP server

说明与用途

构建了一个本地 MCP 服务器，公开四个合规工具：法国电话营销同意检查、按保留会议成本排名的潜在来源、离岸 GDPR 差距报告以及最低成本可用模型选择，支持外呼和离岸团队的可重复、无网络依赖检查。

问题

Small call‑centre outsourcing brokerage needed quick, repeatable checks for French telemarketing consent, lead‑source cost per meeting, GDPR gaps for offshore customer service, and lowest‑cost capable model selection.

结果

The MCP server provides a consistent, network‑free, read‑only interface for compliance checks, enabling rapid, repeatable decision making for outbound and offshore teams.

工作流

• 1. Implement four Python packages: demarchage-consent, cost-per-kept-meeting, rgpd-offshore-check, least-cost-capable.
• 2. Wrap each package into an MCP tool with annotations (LOCAL_READ_ONLY).
• 3. Publish the MCP server to PyPI using Trusted Publishing and to the MCP Registry with ownership verification.
• 4. Configure MCP clients (Claude Desktop, Cursor) to point to the local server via JSON config.
• 5. Test tools through an in‑memory MCP client session to verify names, schemas, and annotations.
• 6. Publish to Glama, ensuring glama.json and Dockerfile use a dedicated virtual environment for uv.
• 7. Deploy the server locally and expose the four tools for repeated use.

证据

Over the last weeks we turned those questions into small, standard‑library Python packages.

来源

This weekend we wrapped four of them in an MCP server, so that Claude, Cursor and other MCP clients can call them directly: kanibal-tools-mcp.

来源

Local, read‑only, closed‑world. Every tool computes from its arguments only: no network call, no file access, no state kept between calls.

来源

The tests do not call the Python functions directly. They go through an in‑memory MCP client session, so they also check the tool names, schemas and annotations that a client sees.

来源