dart-risk-mcp

README.md

# dart-risk-mcp

개인 작업용. DART 공시를 흐름으로 읽어 불공정거래 신호를 사실로 표기하는 MCP 서버.

## 돌리기

```bash
pip install -e .
python -m dart_risk_mcp
```

```json
{
  "mcpServers": {
    "dart-risk": {
      "command": "python",
      "args": ["-m", "dart_risk_mcp"],
      "env": { "DART_API_KEY": "..." }
    }
  }
}
```

- `DART_API_KEY` 필수
- `KRX_API_KEY` · `KIS_APP_KEY` + `KIS_APP_SECRET` 선택 — 시세 대조용(KRX 먼저, 빈 날은 KIS)

## 메모

- 설계·실측 근거는 전부 `CLAUDE.md`
- 보류한 판단은 `docs/DEFERRED-DECISIONS.md`
- 뷰어 소스는 `docs/tool/`
- 테스트는 키 없이: `env -u DART_API_KEY python -m pytest tests/ -q`
- 골든 재생성: `python scripts/regen_goldens.py`
- 릴리스 순서: `CLAUDE.md` 「릴리스 정책」

MIT. 일부 로직은 kreports-dart-mcp(Apache 2.0)에서 이식 — `THIRD_PARTY_NOTICES.md`.

Deploy Server

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A3.7/5.0

Scored 2026-10-27 22:10 across 35 tools

Disambiguation4/5

Most tools carve out distinct scopes and the descriptions explicitly cross-reference neighbours (get_financial_summary vs get_financial_statements_full vs compare_financials; get_audit_opinion_history vs get_audit_opinion_text; get_disclosure_document vs list_disclosure_sections vs view_disclosure). The risk cluster is the soft spot: analyze_company_risk, check_disclosure_anomaly, build_event_timeline and find_risk_precedents all produce 'risk signal' output, and an agent could plausibly pick the wrong one for a given question. get_shareholder_info vs track_insider_trading also overlap on ownership data (snapshot vs time series).

Naming Consistency5/5

All 35 names are snake_case verb_noun (find_, get_, check_, list_, view_, search_, track_, scan_, build_, lookup_, manage_, compare_). Verb variety (track_/scan_/build_) tracks meaningful differences in behavior rather than ad-hoc style drift, and no camelCase or bare-noun names appear.

Tool Count2/5

35 tools is well past the heavy threshold, and several clusters could be consolidated (three disclosure-reading tools, four financial-statement tools, four risk/scoring-style tools, two audit tools). The domain breadth is genuine — disclosures, financials, capital events, audit, market scan, watchlist — but the count forces agents to search a long list and increases selection error.

Completeness4/5

Coverage is unusually deep for the stated purpose: CRUD-ish disclosure retrieval, financials at multiple granularities, capital structure and mezzanine terms, audit history plus quoted text, ownership/insider data, market-wide scanning, market reaction, and unlisted-company financials. Remaining gaps are modest: no free-text disclosure-title search across a company (only presets and singled-out report_name), no consolidated group/ownership-graph view, and find_risk_precedents exposes a lookback_days arg that is documented as unused.

Maintenance

ActivityActive

ResponsivenessNo issues

TDQS

Maintenance

Resources