Korea Stock MCP is an MCP server for Korean stock analysis, intended for use by AI clients. Its goal is to enable AI analysis based on stock price information and disclosure materials through the official APIs of DART (Data Analysis, Retrieval and Transfer System) and KRX (Korea Exchange). The project states in its README that a new service is being prepared that is based on accurate Korean stock data and helps AI provide more reliable answers, but the specific content and pricing model of the new service are not yet provided.
In terms of purpose, this server mainly serves the retrieval and analysis of data related to Korean listed companies, including disclosure search, disclosure data retrieval, financial statement analysis, and stock quote queries.
The main capabilities consist of the following tools: get_disclosure_list is used to search disclosure reports by disclosure type, company, and date; get_corp_code is used to query the unique number, company name, and stock code of companies subject to DART-registered disclosures. It can be queried by company name or stock code. If the company name search fails, it prompts the user to retry using the stock code; get_disclosure is used to parse the original files of disclosure reports through the DART API. For large documents, it returns the table of contents, and section_id can be used to query a specific section; get_financial_statement provides XBRL financial statements of listed corporations and major unlisted corporations, including all accounting item data in periodic reports; get_stock_base_info provides basic information on individual stocks listed on KOSPI, KOSDAQ, and KONEX, such as stock name, code, and market classification; get_stock_trade_info provides daily trading data for individual stocks on KOSPI, KOSDAQ, and KONEX, such as stock price, trading volume, and market capitalization; get_market_type can query the market classification of a stock (KOSPI/KOSDAQ/KONEX) by stock code; get_today_date provides the current date in YYYYMMDD format, making it easier for AI to accurately obtain the date.
In terms of usage, an API KEY must first be applied for. On the DART side, you need to register with OPEN DART, apply for an API KEY on the authentication key application page, and confirm the issued key in the Open API usage status. On the KRX side, you need to register and log in to KRX OPEN API, apply in "My Page → API Authentication Key Application", and after approval, enter the "Service Use → Stocks" menu, and click "API Use Application" for each of the following 6 items: securities daily buy/sell information, KOSDAQ daily buy/sell information, KONEX daily buy/sell information, securities individual stock basic information, KOSDAQ individual stock basic information, and KONEX individual stock basic information. Approval takes about 1 day. After approval, confirm the key in "My Page → API Authentication Key Issuance Details". Then configure it in Claude Desktop: open Claude Desktop, go to "Settings → Developer → Edit Config", add the configuration content to the claude_desktop_config.json file, and restart Claude Desktop for it to take effect. For local development testing, you need to install dependencies, create a .env file in the project root directory to set environment variables, run MCP Inspector, and then open in the browser the address output in the terminal (such as http://localhost:6274），在 In the Inspector, click Connect, then select a tool from the tool list on the left to call it directly; after modifying and saving a file under src/, the server will automatically restart, and clicking Reconnect in the Inspector will immediately reflect the changes. Note that Node.js 18 or later is required.
Regarding the mechanism for handling large disclosure reports, some disclosure reports (such as business reports, quarterly reports, etc.) are large XML documents of several MB, and processing them all at once may exceed AI response limits or cause errors. This tool automatically switches processing methods based on document size: documents smaller than 1MB return the full text directly; documents of 1MB or more return a table of contents (TOC), and Claude automatically selects the sections relevant to the question after viewing the TOC and retrieves their content, with no manual operation required. After answering, it also lists the remaining sections, and the user can continue to request other sections. Users can also directly specify a section in the TOC (e.g., "query section 3"). For complex questions, it is recommended to split them into multiple steps, for example, first obtain the corp_code, then search the disclosure list related to securities issuance, then retrieve the body of that disclosure, and finally look up the details of the foreign investor allotment.
Applicable scenarios include: financial analysis (e.g., querying a company's sales, operating profit, and growth rate for each quarter over multiple years), corporate analysis (e.g., understanding how a company makes money and the sales of each department), disclosure analysis (e.g., querying the details of the foreign investor allotment for an IPO), etc.
Regarding limitations, the project clearly states that this tool is for information provision purposes only, does not constitute investment advice, and all investment decisions are the responsibility of the individual. The API KEY must be applied for by the user, and some KRX interfaces must be applied for individually and require about 1 day for approval. Specific details, fees, and pricing for the new service are not yet provided. The project welcomes contributions via Pull Request, and issues can be registered on GitHub Issues.
