Loading ToolBench
MCP server exposing TOTVS documentation search, Protheus database query, and Confluence integration tools
mapj exposes 8 well-named, read-only tools with complete input/output schemas and tool annotations. Naming follows verb_noun convention (search_totvs_docs, fetch_totvs_doc, protheus_query). All tools have descriptions (avg ~80 chars, within baseline 34 - 392). Input schemas are fully typed with parameter descriptions. However, descriptions are terse and lack actionable context for LLM selection, they state WHAT but not WHEN or WHY. Error handling is absent from visible code; no recovery guidance or categorization. Output schemas are declared but not shown in source, so completeness cannot be verified. Pagination support exists (cursor params) but limits are not documented. No examples of how tools chain together.
Fetch a Confluence page by ID or URL and convert to Markdown
Fetch and convert a TOTVS document to Markdown with metadata and optional pagination
List available TDN Confluence spaces
List available filters for TOTVS documentation search
Describe the schema and optionally sample data from a Protheus table
List tables available in a Protheus database
Descriptions lack actionable context. All 8 tools have 50 - 80 char descriptions stating WHAT but not WHEN to use or prerequisites. E.g., 'Search TOTVS documentation' does not explain when to call search_totvs_docs vs fetch_totvs_doc, or what query formats work. LLMs cannot distinguish between similar tools without explicit guidance.
No error handling or recovery guidance visible in source. Tools accept parameters (e.g., profile_name, query) but no code shown for validation, error categorization, or actionable error messages. If a query fails or profile is not found, LLM receives no guidance on retry vs. user correction.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2025-06-18+ | v2 |
Execute a read-only SQL query against a Protheus database
Search TOTVS documentation across multiple sources including TDN, Zendesk, forum, and web
Output schemas declared but not visible in source code. Schema references (schema.SearchOutputSchema, schema.FetchOutputSchema) are imported from internal/schema package, so completeness, field naming, and chaining IDs cannot be verified.
Result limits not documented. search_totvs_docs accepts max_results param but no description states the default, min, or max. Pagination via cursor exists but guidance on when to paginate is absent.
Tool composition and chaining not documented. No descriptions explain which tools to call in sequence (e.g., 'Call list_totvs_filters first to see available product lines, then pass them to search_totvs_docs'). Agents must infer dependencies, wasting reasoning cycles.