MCP server for Asian financial filings (Japan EDINET, Korea DART) with advanced XBRL analysis
Asia Filings MCP Server has ONE mega-tool (asia-filings) that bundles 18 distinct operations into a single tool with a massive union-type schema. The naming, while descriptive overall, violates critical patterns: (1) The tool name 'asia-filings' is a noun, not a verb, it should be something like 'query-asia-filings' or 'access-asia-filings'. (2) The input schema uses a 'method' enum to switch between 18 entirely different operations, some read-only (search, get) and some potentially state-modifying (though all are documented as READ_ONLY). This violates the single-responsibility principle, each method should be a separate tool. (3) Many parameters are context-dependent (e.g., 'edinet_code' only applies to Japan methods, 'corp_code' only to Korea), making the schema internally inconsistent and forcing the LLM to reason about which parameters to include. (4) The parameter descriptions are present but generic, they state WHAT the parameter is (e.g., 'EDINET code (E-number)') without explaining WHEN to use it or what it controls. (5) While an output schema is not visible in the provided source, the tool description does not document return structure, making it impossible for an LLM to plan downstream actions. The tool description itself is 239 chars and reasonably clear, but it conflates multiple concerns ('company filings, financial statements, and XBRL data') into one sentence. The 'method' enum description is extensive (with good examples), which is a positive, but the parameterization remains fundamentally flawed for an agent-first tool interface.
Unified tool for Asian financial filings: access company filings, financial statements, and XBRL data from Japan (EDINET) and South Korea (DART). Provides comprehensive access to financial reports from 7,700+ Asian companies.
Tool name is a noun ('asia-filings'), not a verb. Names should start with action verbs (get_, search_, query_, access_) so LLMs infer intent from the name alone.
Single mega-tool bundling 18 distinct operations (7 Japan methods, 8 Korea methods, 3 advanced analysis). Each should be a separate tool with a clear verb_noun name (e.g., search_japan_companies, get_korea_financial_statements). This violates single-responsibility and forces the LLM to reason about which 'method' to invoke.
Parameter schema uses context-dependent fields: 'edinet_code' only valid for Japan methods, 'corp_code' only for Korea methods, 'business_year' only for specific Korea methods. LLM must reason about which parameters are valid for each 'method', no schema-level validation prevents invalid combinations like method='search_japan_companies' + corp_code='12345'.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 35 | - | v1 |
No documented output schema. Tool description states it 'Provides comprehensive access to financial reports' but does not explain what fields the response contains, how results are paginated, or what IDs/references are returned for chaining with other tools.
Parameter descriptions are present but generic. They state what a parameter is (e.g., 'EDINET code (E-number)') without dependency hints. No guidance like 'Required only for methods starting with get_japan_*' or 'Use this if you only have a company code, not a filing ID'.
No error handling guidance visible. Tool definition does not document failure modes, recovery paths, or what to do if a company is not found, a filing is unavailable, or an API returns rate-limit errors.
Tool description mentions '7,700+ Asian companies' but does not explain discovery, how does an LLM know which companies are available? No guidance on using search_japan_companies or search_korea_companies first.