MCP Integration — KoreanAds

1499 Ad Products

745 Verified pricing

1238 A/B quality items

63 Media subtypes

📡

What is MCP?

Model Context Protocol (MCP) is an open protocol that enables AI models to access external data and tools. KoreanAds provides 12 professional tools via MCP server, giving AI tools real-time access to Korean ad product databases and the ability to generate expert advertising strategies.

Core use cases

KoreanAds MCP is designed to finish discovery, evaluation, and allocation inside an AI conversation.

Find

Find relevant Korean ad products from a brief and budget using natural language.

Evaluate

Explain why a product fits based on pricing confidence and audience/targeting evidence.

Allocate

Generate an executable media mix using only products with usable price evidence.

From brief to outreach in one workflow

This page is positioned around real planning and sales workflows, not just protocol setup.

Agency planning

Describe target, budget, and KPI goals to the model and let KoreanAds MCP build an evidence-based media shortlist.

Brand buying review

Compare shortlisted products, validate pricing evidence, and prepare outreach copy before talking to publishers.

Global team handoff

Use multilingual UI and structured MCP responses to explain Korean inventory to non-Korean stakeholders.

🛠️ Available Tools

Access 12 professional tools through MCP

🔍 Search & Lookup

🔍 search_ad_products

Search ad products with natural language. Dual vector semantic + SQL fallback search with budget, subtype, and quality score filtering.

query Required — Search query (any language)

media_type Optional — Media type (Digital, TV, OOH, etc.)

media_subtype Optional — Media subtype (Search, Display, Social, Video, etc.)

max_budget_krw Optional — Max budget (KRW)

min_quality_score Optional — Min quality score (0-100)

top_k Optional — Max results (default 5)

📋 get_product_detail

Get complete details for a specific ad product including pricing, targeting, specs, KPI benchmarks and more.

product_id Required — Product ID

⭐ get_top_quality_products

Browse premium ad products sorted by data quality score. Explore what's available without keyword search.

limit Optional — Max results (default 10)

media_type Optional — Media type (Digital, TV, OOH, etc.)

media_subtype Optional — Media subtype (Search, Display, Social, Video, etc.)

🏢 get_media_owner_portfolio

Get the full ad product portfolio of a specific publisher (e.g. Kakao, Naver, Toss) with price ranges.

media_owner_name Required — Publisher name

📊 Analysis & Strategy

🎯 recommend_for_brief

Get DB-grounded recommendations based on a campaign brief. Returns verified product candidates and short planning guidance as grounded JSON.

brief Required — Campaign brief

budget Optional — Budget

media_type Optional — Media type (Digital, TV, OOH, etc.)

media_subtype Optional — Media subtype (Search, Display, Social, Video, etc.)

📊 build_media_mix_from_db

Auto-generate a media mix using only price-verified real ad products. Calculates channel distribution with quality-weighted budget allocation.

brief Required — Campaign description

monthly_budget_krw Required — Monthly budget (KRW)

media_type Optional — Media type (Digital, TV, OOH, etc.)

media_subtype Optional — Media subtype (Search, Display, Social, Video, etc.)

🔄 generate_full_funnel_strategy

Generate a full-funnel strategy (Awareness→Consideration→Conversion) with DB products. Budget split at 40/30/30.

brief Required — Campaign brief

monthly_budget_krw Required — Monthly budget (KRW)

💰 analyze_market_pricing

Analyze market pricing benchmarks by media category. Returns min/avg/max CPM, CPC, CPV stats and budget benchmarks.

media_type Optional — Media type (Digital, TV, OOH, etc.)

media_subtype Optional — Media subtype (Search, Display, Social, Video, etc.)

⚡ Utility

⚖️ compare_products

Compare 2-5 ad products side-by-side on pricing, targeting, billing model, and KPIs.

product_ids Required — Product ID list (2-5)

📈 simulate_campaign_perf

Simulate expected campaign performance for a specific product and budget based on CPM/CPC/CPV unit pricing.

product_id Required — Product ID

budget_krw Required — Budget (KRW)

🧭 check_db_coverage

Routing tool that determines if a query can be answered by the ad DB, or should be handled by the LLM's own knowledge.

query Required — Search query (any language)

📉 get_db_stats

Get DB statistics: total products, media type breakdown, data completeness, and average quality scores.

— No parameters required —

🌐 Endpoints

Choose the appropriate endpoint for your use case

Recommended

Streamable HTTP

https://koreanads.com/mcp

Latest MCP standard. Supported by Claude Desktop, Cursor and most clients.

Legacy

SSE (Server-Sent Events)

https://koreanads.com/sse

SSE-compatible endpoint for legacy MCP clients.

ChatGPT

OpenAPI Actions

https://koreanads.com/openapi-actions.json

A slim OpenAPI schema that Custom GPT Actions can import directly. Use HTTP Actions instead of MCP for ChatGPT.

ChatGPT Custom GPT / HTTP Actions

Use this path when you want a Custom GPT to call KoreanAds over direct HTTP Actions. It is separate from the MCP app path, and it calls search, compare, detail, and media mix endpoints over HTTP.

1

Open Custom GPT

In ChatGPT, open Create a GPT or an existing GPT, then go to the Configure tab and open Actions.

2

Import OpenAPI

Paste the URL below into the Actions schema URL field.

OpenAPI URL

https://koreanads.com/openapi-actions.json

3

Paste Instructions

Paste the following text into Instructions. The default rule is <strong>API first, grounded JSON first</strong>.

Custom GPT Instructions

Use the HTTP API directly.

Primary tools:
- Ad product search: POST /api/tools/search
- Product detail: GET /api/tools/products/{product_id}
- Product comparison: POST /api/tools/compare
- DB stats: GET /api/tools/stats
- Publisher portfolio: GET /api/tools/media-owner-portfolio
- Media mix generation: POST /api/strategy/media-mix
- Brief-based grounded recommendation: POST /api/strategy/recommend

Operating rules:
- Prefer direct JSON tool calls over prose generation whenever possible.
- In MCP, use grounded JSON results only.
- If the user wants to find ad products, call /api/tools/search first.
- If the user wants to compare 2-5 known products, call /api/tools/compare.
- If the user wants details for one product, call /api/tools/products/{product_id}.
- If the user wants a budget split or media plan, call /api/strategy/media-mix.
- If the user wants brief-based recommendations, call /api/strategy/recommend.
- Respond with grounded facts and numbers from the tool result first, then add short explanation only when needed.
- Do not invent product attributes or pricing that are not returned by the API.
- If a product ID is required, find it from search results before calling the next tool.
- Do not write long strategy documents.

4

How to Operate

Use Action JSON results directly for search, compare, detail, stats, and recommendations. Apps and Actions are mutually exclusive inside one GPT. Narrative LLM generation is outside the MCP scope.

⚙️ Setup Guide

Follow the steps for your AI tool of choice

1

Open Config File

Claude Desktop → Settings → Developer → Edit Config

Or edit the file directly:

macOS

~/Library/Application Support/Claude/claude_desktop_config.json

Windows

%APPDATA%\Claude\claude_desktop_config.json

2

Register MCP Server

Add the following JSON to the config file:

claude_desktop_config.json

{
  "mcpServers": {
    "koreanads": {
      "url": "https://koreanads.com/mcp"
    }
  }
}

3

Restart Claude

After saving, restart Claude Desktop and KoreanAds will appear in the tool list.

1

Open MCP Settings

Cursor → Settings → MCP tab

2

Add Server

Click + Add new MCP server and enter the details below:

.cursor/mcp.json

{
  "mcpServers": {
    "koreanads": {
      "url": "https://koreanads.com/mcp"
    }
  }
}

3

Start Using

In Agent mode, use natural language like "Search Korean ad products" to invoke the MCP tools.

1

Connect in ChatGPT

In ChatGPT web, go to Settings → Apps. Availability depends on your plan and workspace permissions. To register your own MCP app, enable developer mode, then use Apps → Create or Workspace Settings → Apps → Create.

2

Enter Server URL

Enter the remote MCP server URL below. ChatGPT does not support local MCP servers, so use a public HTTPS endpoint:

MCP Server URL

https://koreanads.com/mcp

3

Connect and refresh updates

Start a new chat and select KoreanAds from the Apps menu. After an admin approves the app, ChatGPT uses a frozen tool snapshot, so later MCP tool changes require Workspace Settings → Apps → Refresh and then Publish again.

1

Connect via HTTP

Connect programmatically using an MCP client library:

Python (fastmcp)

from fastmcp import Client

async with Client("https://koreanads.com/mcp") as client:
    result = await client.call_tool(
        "search_ad_products",
        {"query": "20대 여성 타겟 SNS 광고"}
    )
    print(result)

TypeScript

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";

const transport = new StreamableHTTPClientTransport(
  new URL("https://koreanads.com/mcp")
);
const client = new Client({ name: "my-app", version: "1.0" });
await client.connect(transport);

const result = await client.callTool({
  name: "search_ad_products",
  arguments: { query: "20대 여성 타겟 SNS 광고" }
});

2

SSE Legacy Method

Use the SSE endpoint for legacy MCP clients:

SSE URL

https://koreanads.com/sse

🚀 Get Started Now

Connect KoreanAds to your AI tools and search, compare, and build media mixes through conversation.

Try Ad Search Media Mix Recommendation

Connect your AI tools to Korean ad data

What is MCP?

Core use cases

Find

Evaluate

Allocate

From brief to outreach in one workflow

Agency planning

Brand buying review

Global team handoff

🛠️ Available Tools

🔍 Search & Lookup

📊 Analysis & Strategy

⚡ Utility

🌐 Endpoints

Streamable HTTP

SSE (Server-Sent Events)

OpenAPI Actions

ChatGPT Custom GPT / HTTP Actions

Open Custom GPT

Import OpenAPI

Paste Instructions

How to Operate

⚙️ Setup Guide

Open Config File

Register MCP Server

Restart Claude

Open MCP Settings

Add Server

Start Using

Connect in ChatGPT

Enter Server URL

Connect and refresh updates

Connect via HTTP

SSE Legacy Method

🚀 Get Started Now

Connect your AI tools
to Korean ad data