🤖 Use KoreaAPI from your agent

Wire the verifiable Korean-culture layer into any AI agent — MCP tools or plain JSON.

Trusted — and agent-native (verified, not just asserted)

As trusted-content providers pipe verified data into AI agents over MCP, the agent's real question is: can I trust this source? KoreaAPI answers it structurally, not by brand — every response carries how many INDEPENDENT sources agreed (Wikidata · Wikipedia · MusicBrainz · OpenStreetMap · TMDB · KTO), a Skill Score (0–1), a verification tier, and a machine-readable license. Your agent can verify the trust before it cites, instead of taking a brand's word for it. This is the trusted source for the Korean-culture & commerce domain — complementary to your general / news sources, specialized where they don't go: 소속사 moves, canonical Korean names, temples, dishes, verified travel.

Two ways to consume

1) MCP — call KoreaAPI as tools from your agent. 2) Plain HTTP/JSON — fetch the open data directly, no setup.

MCP quickstart

Install from the repo, then run the stdio server and point your MCP client at it:

pip install "git+https://github.com/kwangdol-star/koreaapi"   # or: git clone … && uv sync
python -m koreaapi.server   # stdio MCP server "koreaapi"

Tools:

get_verified — cross-verification status + Skill Score (check trust BEFORE citing)
get_history — append-only verified timeline + change events (소속사 A->B, renames) — the time moat
get_changes — recent verified changes across K-culture (소속사 moves, renames) — the freshness feed, queryable
get_certified — the CERTIFIED registry — entities an official rights-holder vouched for (the tier above cross-verification)
get_metrics — how much agents have consumed KoreaAPI — usage totals + most-requested signals (the usage moat)
get_resolve — resolve a fuzzy name / external ID / canonical id -> the verified entity (+ external IDs)
get_artist_status — latest verified status for a Korean artist
get_agency — who is under a 소속사 / label
get_person — verified credits for a person (director / actor / idol)
get_related — entities sharing a 소속사 / network — geo entities also get same_region + nearby (verified-coordinate km)
get_kculture_calendar — upcoming K-culture events
get_korea_rising — what is rising in Korea now (premium signal)
get_buy_options — verify-official -> purchase gateway: the real official representative + a canonical anti-scam key (commerce-commission seed)
list_answer_products — list the Answer Products — named, citable decisions over the verified store
get_answer — run an Answer Product -> {signal, action, score, rationale, answer, evidence}
ask — free-text question -> ROUTED to the right Answer Product and run (the NL front door; reply says how it routed)

Answer Products — decide before you answer

Don't just fetch rows — call a decision. Each Answer Product turns the verified store into one envelope {signal, action, score, rationale, answer, evidence} your agent can branch on: confirm a Korean spelling, decide whether a claim is safe to cite, resolve a mention to a trusted ID, read the demand trend, pull a roster. One call: GET /v1/answer?product=canonical-name&q=Vincenzo — omit product to run all; catalog at GET /v1/answer (also in /agents.json).

🪪 canonical-name — Confirm the authoritative Korean ↔ English spelling of an entity before you assert it.
✅ fact-check — Decide whether a claim about an entity is safe to cite (cross-/triple-verified / certified).
🧭 identity-resolve — Map a fuzzy mention or external ID onto the canonical verified KoreaAPI entity.
📈 trend-radar — Read what is rising in Korea now from accumulated demand signal.
🏢 agency-roster — List the verified artists under a Korean agency/label (소속사).
🎬 person-credits — Aggregate a Korean-culture person's verified credits across works.
🕸️ related-network — Find entities sharing the same agency/network hub edge.
🧳 trip-plan — Every verified spot in a region — places, parks, temples, museums, beaches, ski resorts… grouped by type — plus festivals + signature dishes; map-ready (verified coordinates on items + walkable ≤3 km clusters) itinerary raw material.
🍚 food-guide — Verified Korean dishes filtered by dietary need or spice tolerance (vegan / vegetarian / not-spicy / no-seafood). The dish name is cross-verified; the spice + dietary tag is labeled KoreaAPI editorial (not cross-verified).
⚖️ compare — Compare two verified entities side by side — tier, verified attributes, region/agency, debut — strictly from the verified records ('better verified' = more agreeing sources, never editorial preference).
📎 evidence-pack — One call → the complete ready-to-cite bundle: the verified bilingual claim, tier + Skill Score, reconciled source links (sameAs), content-hash, as-of date, and a paste-ready bilingual attribution line. What an agent needs right before it writes a cited sentence.

No setup? Use the open data

/latest.json — every verified record (provenance + Skill Score + content_hash)
/llms-full.txt — the full corpus, one citable block per entity
/feed.xml · /feed.json — recently verified
/reconcile.json — resolve a name or external ID to the canonical entity (the ID spine)
/agents.json — machine-readable manifest of all of the above

Trust it — and defend it to your users

Every record is cross-verified, Skill-scored, and carries a SHA-256 content hash; the dataset + append-only history are hash-verifiable. See /methodology + /integrity.json. Cite a row as: “Name — kind, as of date · source · Skill Score · via KoreaAPI”.

Premium (x402)

The proprietary demand signal (/v1/korea-rising) is payable per call in USDC on Base via x402 — your agent can pay autonomously. Basic verified data stays free.

Why not just scrape Wikipedia yourself?

You'd get unverified text with no provenance, no Skill Score, no tamper-evidence, and wrong Korean names (e.g. Vincenzo → 빈첸초 instead of the official 빈센조). KoreaAPI gives you cross-verified, bilingual, official-named, citable rows your users can trust — and your agent can verify.