Disclosure Compass
Integrates with the OpenDART public disclosure system to retrieve company profiles, financial statements, dividend information, major shareholders, employee statistics, and search/filter disclosure filings.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Disclosure CompassShow me recent disclosures from Samsung Electronics"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Disclosure Compass (공시나침반)
A focused Model Context Protocol server for public company disclosures from OpenDART (전자공시시스템 DART). Disclosure Compass (공시나침반) exposes ten read-only gateway tools over Streamable HTTP. Behind that compact public surface are 16 in-process specialist MCP servers with 82 OpenDART tools, distilled from the original implementation without its LLM, vector-search, or A2A runtime dependencies.
This project is independent open-source software and is not an official OpenDART product.
Tools
Tool | Purpose |
| Basic company and listing information |
| Disclosure filings in a bounded date range |
| A bounded set of statement accounts |
| Current and prior-period dividend rows |
| Major shareholder positions |
| Employee counts, tenure, and pay statistics |
| Rank a Korean request across 16 disclosure domains and show their specialist tools |
| Inspect all 16 specialist servers and 82 tool names/endpoints |
| Execute an exact named tool on an exact specialist server |
| Classify, select a specialist tool, and execute it in one call |
All tools are declared read-only, non-destructive, idempotent, and open-world. Every specialist result is capped at 50 rows. Binary filing and XBRL tools return bounded ZIP metadata rather than raw file contents. Upstream calls use strict connect/read timeouts and reject JSON responses larger than 2 MB or binary responses larger than 8 MB.
Related MCP server: opendart-fss-mcp
Disclosure request classification
classify_disclosure_request maps a non-empty Korean natural-language request
of up to 500 characters to the most relevant OpenDART disclosure domains.
top_k defaults to 3 and may be set from 1 through 16. The response contains
the original query, total_categories, and score-ranked routes. Each route
contains:
category: the stable domain IDlabel_ko: the Korean domain labelscore: the normalized relevance scorematched_terms: terms from the request that contributed to the routesupported_tools: data tools currently exposed by this server for the domainspecialist_tools: original tool names available on that domain's MCP server
This classifier performs deterministic term matching and does not itself call
OpenDART. Use route_and_call_disclosure to classify and execute in one request,
or use call_disclosure_server_tool for deterministic server/tool selection.
The 16 specialist servers are real FastMCP instances called through in-memory
FastMCP clients; they are not 16 separately deployed network services.
The classifier recognizes exactly these 16 domains:
Category | Korean label |
| 공시검색/기업개황 |
| 주주/주식 정보 |
| 임원/보수 정보 |
| 채무증권 정보 |
| 감사/자금 정보 |
| 재무정보 |
| 지분공시 |
| 증자감자 |
| 자기주식 |
| 전환증권 |
| 합병분할 |
| 영업/자산 양수도 |
| 해외상장 |
| 지분거래 |
| 기업이슈 |
| 증권발행/등록정보 |
Example request:
{
"query": "연결 재무제표 매출액",
"top_k": 1
}Example response:
{
"query": "연결 재무제표 매출액",
"routes": [
{
"category": "financial_statement",
"label_ko": "재무정보",
"score": 1.0,
"matched_terms": ["연결 재무제표", "재무제표", "매출액"],
"supported_tools": ["get_financial_statement"],
"specialist_tools": [
"dart_fnlttSinglAcnt",
"dart_fnlttMultiAcnt",
"dart_fnlttXbrl",
"dart_fnlttSinglAcntAll",
"dart_xbrlTaxonomy",
"dart_fnlttSinglIndx",
"dart_fnlttCmpnyIndx"
]
}
],
"total_categories": 16
}To execute a routed request directly, pass a company identifier and any
period-specific fields to route_and_call_disclosure:
{
"query": "삼성전자 전환사채 발행 결정을 찾아줘",
"corp_code": "00126380",
"begin_date": "20250101",
"end_date": "20251231"
}This selects the convertible_securities server and its
dart_cvbdIsDecsn tool. Callers that already know the exact target can instead
use call_disclosure_server_tool with server_id, tool_name, and an
arguments object. Accepted fields depend on the endpoint and include
corp_code or corp_name, business_year, report_code, begin_date,
end_date, corp_codes, receipt_number, fs_division, statement_type,
index_code, and max_items.
Requirements
Python 3.11-3.13
An OpenDART API key from https://opendart.fss.or.kr/
Run locally
python -m venv .venv
. .venv/bin/activate
pip install -e '.[dev]'
export DART_API_KEY='your-runtime-secret'
opendart-mcpThe MCP endpoint is http://localhost:8000/mcp and the health endpoint is
http://localhost:8000/health. Set PORT to override port 8000.
The corp_code arguments are OpenDART eight-digit company identifiers, not
six-digit stock codes. For example, OpenDART's public guide uses 00126380 for
Samsung Electronics.
Docker / PlayMCP deployment
The image is compatible with Linux AMD64 and runs as a non-root user.
docker build --platform linux/amd64 -t opendart-mcp .
docker run --rm -p 8000:8000 \
-e DART_API_KEY='your-runtime-secret' \
opendart-mcpConfigure the deployment secret as DART_API_KEY; never bake it into the image
or commit it. Register /mcp as the Streamable HTTP endpoint. The server uses
stateless HTTP so it does not require session affinity.
Development checks
ruff check .
pytest -q
python -m compileall -q src testsTests use local fixtures and never call OpenDART or require an API key.
Data and operational notes
Results are public disclosure data, but users should confirm material facts in the original filing before making decisions.
OpenDART may update data between otherwise identical calls.
The server does not store query inputs or results.
Upstream rate limits and maintenance windows still apply.
License
MIT. See LICENSE.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/MLOpsEngineer/opendart-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server