What Data Can It Query?
From the national law information center: 21 tool families covering statutes, court precedents, Constitutional Court decisions, statutory interpretations, administrative rules, local ordinances, and treaties — including specialist tools like three-way comparison (act / decree / regulation) and old-vs-new clause comparison for amendments.
From DART: five tools for disclosure filings, company profiles, full financial statements, and key accounts like revenue and operating profit. From the public data portal: eight everyday datasets — pharmacies, hospitals, veterinary clinics, stock dividends, rare medicines, health supplements, and business-registration verification. For comparative law research, US (CourtListener) and German (OpenLegalData) case search is included.
Why It Was Built — the Problem of Scattered Public Data
Korean public data is already open. The problem is that every agency has different API key procedures, authentication, and response formats. Querying a statute meant learning the legislation API; financials meant DART; pharmacies meant the data portal. Wrapping these behind one MCP layer lets AI call every source the same way.
How to Start — One URL, One Minute
Register a single connector URL under Settings → Connectors in the Claude app and you're done — no API keys, no installation. Claude Desktop and Cursor take one line in a config file; it also connects as a ChatGPT custom GPT. A minute later, "find Supreme Court precedents on unfair dismissal" just works.
Where It's Actually Used
Legal work leads. In a law-firm assistant system, the MCP verifies statute and precedent citations in AI-drafted documents — a safety net against fabricated case numbers going out the door. For corporate analysis it pulls financial statements and disclosures; for transactions it verifies a counterparty's business registration in one chat line.
Public data was always there — just behind walls of API keys and documentation. One AI-readable wrapper turns it into everyone's data. What SH Consulting does in AX is essentially the same job: opening the data an organization already has into a form AI can use.