Bodacc

Several tool families blur together: ask_pipeworx and ask_pipeworx_beta are explicitly described as functionally identical right now, and ask_pipeworx_grounded/deep_research overlap with the same router. The prediction-market family (bet_research, polymarket_edges, polymarket_arbitrage, polymarket_edge_tracker, polymarket_fill_risk, polymarket_kalshi_spread) has heavily overlapping purposes, and ai_visibility_check vs scan_competitor_ai_presence is a wrapper relationship. An agent could easily route a query to the wrong member of these clusters.

Naming is a mix of conventions: verb-first (ask_pipeworx, discover_tools, validate_claim), noun-first (bodacc_search, bodacc_get_notice, bet_research, release_calendar_markets), noun-only (entity_profile, company_facts, recent_alerts), and inconsistent verb pairs (subscribe/unsubscribe next to list_subscriptions). All snake_case, but the verb position and style vary unpredictably across the set.

40 tools is heavy for any single server, and it is a severe mismatch for a server named 'Bodacc' whose actual domain (French business registry notices) is served by only 4 of the 40 tools. The remaining 36 are an unrelated grab-bag spanning SEC filings, prediction markets, npm packages, memory, and AI marketing, with several near-duplicate families inflating the count.

The BODACC core is actually well-covered: search, get-by-id, company event history, and an insolvency monitoring feed form a complete read-only surface. The broader data surface uses meta-routers (ask_pipeworx, deep_research) to paper over gaps, but the sprawl creates dead ends — e.g., subscription feeds exist for sec_8k/fred/polymarket but not for BODACC, and several analysis tools reference underlying market-fetch tools (polymarket_market, kalshi_market) that aren't directly exposed.