Carbon Interface

Several tools have heavily overlapping purposes: ask_pipeworx, ask_pipeworx_beta, and ask_pipeworx_grounded are variants of the same router with only subtle behavioral differences, and polymarket_edges, polymarket_arbitrage, and polymarket_kalshi_spread all scan for trading opportunities with partially overlapping model families and detection logic. Descriptions do explain the distinctions, but an agent could easily misselect among these clusters.

Naming mixes several conventions: snake_case verbs (ask_pipeworx, compare_entities, resolve_entity), prefixed namespacing (polymarket_*, pipeworx_*), and some standalone nouns (recall, remember, forget). It is readable but lacks a single predictable pattern, and the co-existence of ask_pipeworx / ask_pipeworx_beta / ask_pipeworx_grounded alongside differently-formed neighbors makes the scheme feel ad hoc.

34 tools is heavy for a single MCP server, and the surface is clearly a bundle of distinct products (Pipeworx data router, Polymarket betting analytics, carbon estimates, memory, subscriptions, meta-tools). The count reflects an aggregation of unrelated capabilities rather than one coherent, well-scoped domain.

Within several sub-domains coverage is quite thorough: Polymarket has research, edges, arbitrage, fill-risk, tracker, and Kalshi-spread; subscriptions have subscribe/unsubscribe/list/recent_alerts. But carbon only offers three estimate tools with no retrieval, and the overall surface has no clear single domain to fully satisfy, leaving the set feeling broad rather than complete.