Loading ToolBench

Reading the current corpus…

maple-procure MCP server · C · ToolBench

← Browse

maple-procure

github.com/samarpassey/maple-procure

An MCP server over Canada's public tender and award data (CanadaBuys)

2026-07-28+STDIO4 tools

62/100

Definition Quality80

Protocol Readiness47

Supportability41

Spec posture

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

Static source inference · medium confidence · detected: stateful session

Deprecated protocol patterns detected

Summary

MapleProcure demonstrates solid tool design with clear naming (verb_noun pattern), comprehensive parameter descriptions, and well-structured schemas. All four tools have explicit descriptions (194 - 250 chars, within baseline 34 - 392 range) and typed input schemas. Parameters include enums (format, group_by, currency) and constraints (limit 1 - 100). However, output schemas are not documented in the visible code, LLMs cannot predict response structure. The export_tenders tool uses a two-phase confirmation pattern (preview/confirm) which is sophisticated but underdocumented. Error handling and recovery guidance are absent from tool descriptions. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles (3 READ_ONLY, 1 WRITE).

Tools (4)

export_tenderswriteauth50/100

Export tender notices to a file (markdown or CSV) — preview first, then confirm

search_awardsread onlyauthsource verified85/100

Search award notices — who won what, for how much

search_tendersread onlyauth50/100

Search open tender notices — what is out for bid

summarize_spendread onlyauthsource verified81/100

Total disclosed contract value by buyer, supplier or category, for one currency

Issues (5)

HIGHtool ↗

Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). Tools return structured data but descriptions omit what fields to expect.

search_tenderssearch_awardssummarize_spendexport_tenders

HIGHtool-annotations ↗

No tool annotations (readOnlyHint/destructiveHint). Three tools are READ_ONLY, one is WRITE, but MCP annotations are absent. Agents cannot distinguish safe from risky operations.

search_tenderssearch_awardssummarize_spendexport_tenders

MEDIUMconfirmation-request ↗

export_tenders confirmation flow underdocumented. The two-phase preview/confirm pattern (confirmation parameter) is sophisticated but lacks guidance on when to call preview vs confirm, what the handle contains, or error recovery.

export_tenders

MEDIUMrecovery-guide ↗

Recommendations

Add output schema documentation to all tools. Example: 'Returns array of {notice_id, title, description, closing_date, notice_url, notes}. Limit 20 results; use pagination if available.'
Annotate tools with MCP hints: search_tenders/search_awards/summarize_spend get readOnlyHint=true; export_tenders gets destructiveHint=true (writes to disk).
Expand export_tenders description to explain the two-phase flow: 'Call once with tender_refs and format to preview; response includes confirmation handle. Call again with same params + confirmation handle to execute export.'
Add error recovery hints to search tools: 'If no results, try broader keywords, remove date filters, or check region spelling against CanadaBuys categories.'
Document default values in schema: limit defaults to 20; top_n defaults to 10. Specify that date parameters are optional and filter inclusively (closing_before is ≤, not <).

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained

Score history

Overall score trend

First recorded score · v2 rubric

62/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	C	62	2026-07-28+	v2