BYOA Documentation — Last Full Review: 14 August 2026
LuxGov

AI Boundary Guide

What LuxGov's AI access will — and will never — do

The core rule

LuxGov's product is an evidence checklist, not a certification authority. Only a licensed reviewer or certifier makes a compliance determination. Every AI-touched or evidence-derived response — including everything your connected AI shows you — carries this boundary, not just LuxGov's own dashboard.

"This response is a factual, evidence-only extract from LuxGov for a project that already exists in this account. It is not a compliance verdict, certification, legal opinion, or approval of any kind."

This is the exact boundary statement LuxGov's MCP API returns with every response.

Try it yourself

Ask your connected AI something like: "Is this project compliant?" or "Can you certify this project passed?" A correctly working connection will decline to answer with a pass/fail/certified verdict, and instead point you to the underlying evidence and note that a licensed reviewer makes that call. If your AI instead gives you a confident yes/no compliance verdict, it's not drawing that answer from LuxGov's data — treat it with real skepticism.

Supported vs. unsupported AI clients

ClientStatusWhy
Microsoft Copilot StudioSupported — most testedCustom connector import (Swagger 2.0) works against LuxGov's REST API
Custom GPT (ChatGPT Actions)SupportedActions accept a REST/OpenAPI-described API directly
Generic script or automation tool (n8n, Zapier-style)Technically compatible, not formally vettedAny tool that can call a REST API with a header can connect, but LuxGov hasn't tested every platform
Platforms requiring a native MCP connector protocolNot currently supported in Phase 1Requires MCP protocol capabilities not provided by LuxGov's current REST-based connection

Known clients this currently affects: ChatGPT's built-in/native connector feature, and Claude Desktop's native connector feature. Both require the real MCP protocol (JSON-RPC over a persistent connection), which LuxGov's Phase 1 API doesn't implement.

Why "REST API" and "MCP protocol" aren't the same thing

LuxGov's current AI-access layer is a normal, authenticated REST API — the same kind of thing a Custom GPT Action or a Copilot Studio connector already knows how to call. The real Model Context Protocol is a different, stricter specification (JSON-RPC based) that certain platforms' native connector features expect (see the compatibility note above). Building a real MCP-protocol server is possible in the future, but it isn't what exists today — this page will be updated if that changes.

Sovereignty in transit

Once your AI assistant receives information from LuxGov, it leaves LuxGov's sovereign storage environment and is subject to your AI provider's own data handling. If your organization has jurisdiction-specific data residency obligations beyond what LuxGov's storage provides, confirm your AI provider's processing location before connecting.

What this connection will never do

Not claimed, not built (as of 14 August 2026)
  • Never issues a compliance verdict, pass/fail, "approved," or "certified" label
  • Never uploads documents, triggers analysis, or modifies anything in your LuxGov account
  • Never sees your original question or your AI's final answer — only which tool was called and with what parameters
  • Does not support platforms that require a native MCP connector protocol today (see the compatibility table above)
  • Does not yet detect conflicts between documents (an honest "not available" response, not a fabricated one)