An MCP server for accessing financial reports, company information, and related data from the Financial Reports API
The server exposes 15 tools with READ_ONLY risk classification. However, critical quality gaps significantly impact usability: (1) Tools 13 - 15 (search_companies, get_company_detail, get_latest_filings) have inferred definitions visible only in example code, not in explicit server registration. (2) Parameter descriptions are present but generic (e.g., 'The industry code' for industry_code). (3) No output schemas are documented, LLMs cannot predict response structure, limiting multi-step planning. (4) Descriptions lack context on WHEN to use a tool vs alternatives (e.g., search_companies vs list_industries). (5) Complex nested objects in params (e.g., search_companies.params with conditional subfields) lack clear dependency documentation. (6) No error recovery guidance, tools return raw API responses without actionable error messages. The naming is consistent (verb_noun pattern: get_, list_, search_) and tool names are clear, but descriptions fall short of the 50 - 200 char LLM-optimized range, averaging ~40 chars. This is a functional but incomplete implementation.
Get detailed information about a company by its unique numeric ID.
Get detailed information about a filing type by its ID.
Get detailed information about a GICS industry by its code.
Get detailed information about a GICS industry group by its code.
Get the latest financial filings, optionally filtered by company, ISIN, type, language, etc.
Get processed content for a filing by its ProcessedFiling ID.
Tools 13 - 15 (search_companies, get_company_detail, get_latest_filings) are defined only in example code (examples/python_client_example.py), not in explicit server registration.
No output schemas documented for any tool. LLMs cannot predict response structure, cannot plan multi-step chains, and cannot extract required fields for downstream calls. Example: after calling search_companies, does it return company_id or company.id? This forces wasteful discovery calls.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | F | 41 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 46 | - | v1 |
Get the OpenAPI3 schema for the Financial Reports API.
Get detailed information about a GICS sector by its code.
Get detailed information about a data source by its ID.
Get detailed information about a GICS sub-industry by its code.
List all available GICS industries, optionally filtered by industry group code.
List all available GICS industry groups, optionally filtered by sector.
List all available data sources.
List all available GICS sub-industries, optionally filtered by industry code.
Search for companies by name, ISIN, or LEI, with advanced filtering.
Descriptions are generic and too short (avg ~40 chars; baseline 194 chars for A-grade tools). Examples: 'Get detailed information about a filing type by its ID' (51 chars) lacks WHEN to use it vs list_filing_types. Add context: e.g., 'Retrieve metadata for a GICS industry (e.g., code "1010"); call list_industries first if unsure of the code.'
Parameter descriptions lack dependency and constraint information. Example: search_companies.params object contains sub-fields (sector, industry_group, industry, sub_industry) with no statement that these are mutually exclusive or dependent (GICS hierarchy: sector → industry_group → industry → sub_industry). Document: 'Filters are hierarchical; specifying sub_industry without industry is invalid.'
No error recovery guidance. Tools return raw API responses without actionable error messages. E.g., if search_companies returns 0 results, does it say 'No companies found matching filters (try relaxing sector/industry filters)' or a bare 404? Agents need recovery hints.
Pagination parameters (page, page_size) lack limits and guidance. E.g., list_industries accepts page_size with default 100 but no stated maximum. Add: 'page_size: integer, 1 - 100 (default 10).'
Inconsistent parameter naming. Tools use 'industry_group_code' (list_industries) vs 'group_code' (get_industry_group); 'sub_industry_code' vs 'industry_code'.
Complex nested parameter structure (search_companies.params, get_latest_filings.params are object types with multiple sub-fields) not validated or documented. LLMs struggle with deeply nested inputs. Consider flattening to search_companies(search, countries, sector, industry_group, industry, sub_industry, page, page_size) or document the exact JSON structure required.
get_schema tool is a meta-tool that returns the API OpenAPI spec. Unclear WHEN an LLM should call this vs the other 14 tools. Add context: 'Call this if you need to explore the full API schema; most common queries are already covered by the dedicated tools above.'