Arcgis Detroit

The server mixes two very different domains: Detroit GIS (layer_info, query_layer, search_datasets — distinguished fine) and the massive Pipeworx data platform, where several tools overlap heavily: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, and deep_research all route questions to the same 5,957 sources with only nuanced differences. A user asking a factual question could reasonably pick any of the four, and bet_research vs polymarket_edges vs polymarket_arbitrage also overlap on Polymarket market analysis. The descriptions are extremely verbose and try to disambiguate via use-case, which helps, but the boundaries remain fuzzy.

Naming is wildly inconsistent. Most tools use snake_case verb_noun (search_datasets, query_layer, layer_info, resolve_entity) but several are bare nouns (recall, remember, forget) and others mix domains and styles: camelCase-ish 'generate_llms_txt', prefix-heavy 'ask_pipeworx_beta', 'polymarket_kalshi_spread', 'ai_visibility_check', 'scan_competitor_ai_presence'. There is no single predictable pattern across the full set.

34 tools is heavy for what is ostensibly a Detroit GIS server, and the count is inflated by the Pipeworx platform surface. Many tools (ask_pipeworx family, polymarket_* suite, memory trio) are clearly part of a separate, very broad platform rather than the stated Detroit GIS purpose. This feels like two servers glued together, over-scoped for the niche.

For the Detroit GIS niche, the three tools (search_datasets, layer_info, query_layer) provide discovery and reading over feature services but lack any write/update/delete or metadata management, so the local surface is thin. The broader Pipeworx surface is enormous and covers many domains, but heavily skewed toward financial/prediction data; other domains (generic file storage, user management, mapping operations beyond feature queries) are absent.