Loading ToolBench

Reading the current corpus…

inpi-marcas-mcp-server MCP server · D · ToolBench

inpi-marcas-mcp-server

github.com/milaniseb12-png/inpi-marcas-mcp-server

MCP server for the official Brazilian INPI trademark search system (pePI) — search process numbers, marks, owners and figurative codes with your own INPI login.

2026-07-28+STDIO7 tools

59/100

Definition Quality77

Protocol Readiness45

Supportability38

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

Server has 7 tools with complete input schemas and descriptions. Naming follows verb_noun pattern (search_*, get_*, next_page). Descriptions are present and moderately detailed (100-300 chars typical). However, parameter descriptions are repetitive across tools (salvar_html, forcar_atualizacao, salvar_evidencia appear identically in all 7 tools), suggesting copy-paste rather than thoughtful design. Output schemas are not explicitly documented in the code, responses are inferred from parser functions. Error handling is minimal; no recovery guidance or actionable error messages visible. Security: credentials injected via env vars (good), but no permission gates or audit logging. Composition is reasonable, each tool has a single responsibility, but pagination coupling (lastLiveSearchKey/lastServedSearchKey tracking) introduces subtle state management that complicates the stateless MCP model.

Tools (7)

inpi_get_process_detailread onlyauthsource verified77/100

Obtém detalhes completos de um processo/pedido de marca (situação, datas, classes, titulares, petições, prioridade unionista, etc.)

inpi_next_pageread onlyauthsource verified68/100

Pagina para a próxima página de resultados da busca anterior

inpi_search_advancedread onlyauth50/100

Busca avançada com múltiplos critérios (marca, classe, apresentação, natureza, precisão)

inpi_search_by_markread onlyauthsource verified80/100

Busca por marca (nome, exata ou aproximada)

inpi_search_by_ownerread onlyauthsource verified80/100

Busca por titular (pessoa física ou jurídica) — CNPJ/CPF ou nome

inpi_search_by_processread onlyauth50/100

Busca por número (processo, GRU, protocolo ou registro internacional)

inpi_search_by_vienna_coderead onlyauth

Issues (5)

HIGHtool ↗

Output schemas not documented. Parser functions (parseProcessDetail, parseSearchResults) exist but their return types are not visible in tool definitions. LLMs cannot plan downstream calls without knowing what fields to expect.

inpi_search_by_processinpi_search_by_markinpi_search_by_ownerinpi_search_by_vienna_codeinpi_search_advancedinpi_next_pageinpi_get_process_detail

MEDIUMtool-description ↗

Repetitive parameter descriptions. salvar_html, forcar_atualizacao, and salvar_evidencia are identical across all 7 tools. This suggests copy-paste and reduces clarity about tool-specific behavior. Each tool should explain why these options matter in its context.

inpi_search_by_processinpi_search_by_markinpi_search_by_ownerinpi_search_by_vienna_codeinpi_search_advancedinpi_next_pageinpi_get_process_detail

HIGHrecovery-guide ↗

Recommendations

Document output schemas for all 7 tools. Add a 'Returns' section to each tool description listing key fields (e.g., 'Returns: {results: [{codPedido, marca, titular, status}], totalCount, nextPage}').
Refactor pagination state. Instead of server-side tracking, require inpi_next_page to accept a 'search_context' or 'session_token' parameter that encodes the prior search. This makes each request self-contained and stateless.
Add error handling examples. Document what happens on auth failure, no results, invalid parameters, and API timeouts. Provide recovery hints: 'If auth fails, verify INPI_USERNAME and INPI_PASSWORD. If no results, try busca_exata=false or refine marca.'
Consolidate repetitive parameters. Create a shared 'OutputOptions' schema for salvar_html, forcar_atualizacao, salvar_evidencia and reference it in all tools, reducing duplication and improving maintainability.
Clarify inpi_next_page dependency. Update description to: 'Paginate the most recent search (from inpi_search_by_process, inpi_search_by_mark, etc.). Requires a prior search in the same session.' Consider adding a 'search_tool' parameter to disambiguate.

Score history

Overall score trend

First recorded score · v2 rubric

59/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	D	59	2026-07-28+	v2