fetch
ChatGPTUse after search or search_articles when the user needs full text, a detailed summary, a quotation, a comparison, or deeper analysis. Retrieves one article by UUID and returns id, title, text, URL, and available article metadata, including publication and distribution dates, source, author, publisher, advert value, media reach, and duration in minutes. Zero advert value or media reach is omitted as missing information. Fails when the id is missing or the article is not found.
find_similar_articles
ChatGPTUse when the user wants more media coverage like an article already found. Takes an article UUID returned by a search tool and excludes that source article. Returns count, total, and similar article records with metadata, URLs, summaries, excerpts, authors, source details, duration in minutes, advert value, media reach, article tags, and the article topic-tag subset. Source contains id, name, section, publisher, and tags. Section contains id and title; publisher contains id, country code, and name; source and article tag collections contain id, title, and type. Zero advert value or media reach is omitted as missing information. Authors contain id and name. Call fetch for the full text of any selected result.
read_servanta_skill
ChatGPTUse when complete Servanta workflow guidance is needed and the host or connector does not expose MCP resources. Returns the exact shared Servanta SKILL.md as Markdown. Do not call when the host has already loaded the skill instructions.
search
ChatGPTUse for a simple keyword lookup across the licensed Servanta media-monitoring archive, including standard ChatGPT company-knowledge search. Returns results containing id, title, and URL only. Use search_articles instead for filters, date ranges, pagination, semantic search, or a request for "today"; interpret an unqualified "today" using the distributed date and tell the user that it was used. Call fetch with a returned id before quoting, summarizing, comparing, or analyzing an article.
search_articles
ChatGPTUse when the user needs filtered, date-bounded, paginated, exact-keyword, or semantic search across the licensed Servanta media-monitoring archive. The published date is when the source published or broadcast the item; the distributed date is when Servanta delivered the monitored item into its archive. For an unqualified request for "today", filter by distributed date; use published date when the user explicitly asks what was published today. Always tell the user which date field was used. Result records expose both publication and distribution timestamps so answers can display the date that matches the selected filter. Keyword syntax supports OR alternatives, +required, -excluded, quoted multi-word phrases and wildcard*. Semantic mode embeds the query server-side and ranks articles by meaning. Topic and author UUID filters each match articles containing any supplied UUID and work in either search mode. The default page contains 100 articles. When count equals the requested size, more articles may exist; page with from, but never let from + size exceed 10000. Narrow the date range instead of crossing that limit. Returns count, total, and article records with ids, titles, authors, publication metadata, canonical URLs, summaries, excerpts, source details, duration in minutes, advert value, media reach, article tags, and the article topic-tag subset. Source contains id, name, section, publisher, and tags. Section contains id and title; publisher contains id, country code, and name; source and article tag collections contain id, title, and type. Authors contain id and name. Zero advert value or media reach is omitted as missing information. Call fetch before a detailed summary, quotation, or comparison.