What MCP is, and why connect it to public data
MCP (Model Context Protocol) is a standard channel through which AI models connect to external data and tools. It lets an AI like Claude call a defined data source directly instead of scraping the web.
Korea's public data is vast and open on data.go.kr, but auth-key issuance, parameter specs, and XML parsing make it hard for an AI to use as-is. K Public Data MCP wraps many public APIs (statute law, DART filings, HIRA, and more) into intent-level tools an AI can call, removing that barrier.
Open data an AI can't read is effectively closed. Opening it is less about publishing than about connection.
How search flows into detail
Before, you could only search hospitals by name, region, or type (tertiary, long-term-care, etc.). What equipment a hospital held, or how many specialists worked per department, meant going back to the HIRA site.
This extension uses the encrypted institution code (ykiho) that comes with each search result as the key to detail. Find a hospital and its unique key comes with it; call the detail service with that key and you get five branches at once: facilities, details, departments, equipment, and transit.
1) Search: search_hospital(yadmNm="Samsung Medical Center")
-> result includes ykiho (encrypted institution code)
2) Detail: get_hospital_detail(ykiho)
-> facilities / specialists per dept / equipment / transit at onceWhat actually comes back - Samsung Medical Center
Verified against live data. Point at one hospital and its internal scale shows up in numbers.
Specialists per department expose staffing at the department level; equipment lists the count of high-cost machines like PET, CT, and incubators. Facility info carries the bed configuration.
| Section | Result |
|---|---|
| Facilities | 1 record (beds and facilities) |
| Departments | 29 depts (213 internal-medicine specialists) |
| Equipment | 16 types (4 PET scanners, etc.) |
| Transit | Per-institution registered data |
Who uses this
In trade-area analysis, a hospital's scale and department mix in a given area become inputs for reading medical demand. In medical research or institution comparison, the AI takes over the site-by-site digging a person used to do.
It's all open source. Register a HIRA public-data developer account (no cost) and anyone can wire it into their own AI. The original repo is linked below.