Concylium

Callable expertise, where every answer carries the source it came from.

This instance serves the published work of Nabeel Khan on AI governance and model risk in Canada and the GCC. Every response returns data, citations, caveats and provenance, traced to a published book whose appendices were rebuilt after six verification passes.

Where a regulator has issued nothing, it says so rather than describing a plausible instrument. It cannot quote a price, send anything, or commit anyone to work.

Connect

1. Add the server to your MCP client. The transport is streamable HTTP and there is no credential to obtain.

   {
     "mcpServers": {
       "concylium": {
         "type": "http",
         "url": "https://mcp.nabeelkhan.com/api/mcp"
       }
     }
   }

2. In Claude Code, the same thing in one line.

   claude mcp add --transport http concylium https://mcp.nabeelkhan.com/api/mcp

3. Restart the client and confirm 13 tools appear: get_expert_profile, search_knowledge, lookup_regulation, get_case_study, start_assessment, score_assessment, identify_relevant_service, list_consultation_slots, book_consultation, explain_this_setup, send_message, normalize_arabic_text, get_framework.
4. Ask something a general model gets wrong. lookup_regulation for Saudi Arabia returns four instruments and no banking AI directive, because none has been issued. That negative result is the point of the server.

Machine surfaces

Server card at /.well-known/mcp/server-card.json. The corpus this answers from is published whole at companion-data.json, and the site’s other machine-readable resources are indexed at /docs.