Iskat — Russian-language web search for agents

Each tool targets a distinct operation: core search, batch search, product search, passage-level retrieval, similarity lookup, page fetching, watch management, and index introspection. The potentially overlapping tools are explicitly differentiated, such as search_products versus search, and the batch variants are clearly scoped.

Most tools follow a clear verb_noun snake_case pattern like fetch_page, delete_watch, list_sites, and watch_query. The pattern is slightly weakened by noun-phrase names like index_stats and source_profile, but the naming remains predictable overall.

Fourteen tools is within the ideal range for a search-focused server, and each tool covers a distinct feature: search, batch retrieval, product search, passage extraction, similarity, full-text fetching, watch lifecycle, and corpus introspection. No tool feels redundant or extraneous.

The toolset covers the full agent workflow for web search: querying, reading indexed pages, extracting citable passages, finding related content, monitoring with watches, and understanding corpus coverage and source trustworthiness. Minor conveniences like updating a watch are absent, but the lifecycle is otherwise complete.