edufeed.org

nope-mcp is a search gateway to edufeed's open educational resources on Nostr: learning resources described with AMB metadata — a schema.org/LRMI-based profile for educational content — plus long-form articles, wiki pages, scientific publications, and educational calendar events. The default relay alone holds thousands of learning resources, and the optional extra relays (see list_relays) add tens of thousands more, for well over 100,000 resources across all relays combined. Search results for learning resources are limited to open licenses (CC0, the Public Domain Mark, CC BY, CC BY-SA); articles, wikis, publications, projects, measures and calendar events carry no license filter and are returned regardless of license. Most requests fall into one of two patterns: finding materials to browse, or getting a direct answer to a question with citations drawn from resources' fulltext — the sections below route each case to the right tool. Use these tools to answer questions about educational content, its authors, and upcoming events rather than reading the relays directly; identifiers like naddr, npub, and pubkey are NIP-19/Nostr values these tools return, so pass them back as-is rather than constructing them yourself. The routing guidance below gives example phrases in German, the language of edufeed's primary user base, but the same tool-call patterns apply to requests in any language. Two flows cover most questions: - By name ("materials or events by Jörg Lohrer"): call resolve_author(name) to turn a person or organisation name into pubkey candidates, then pass the chosen pubkey to search_content (and/or search_calendar_events) as authors:[pubkey]. - By topic ("materials on peace education"): call search_content, then hand a result's naddr to get_resource for full metadata. Route by INTENT before picking a search tool (when search_passages is available): - QUESTION intent — the user asks something and wants an answer ("wie kann ich Studierende aktivieren?", "was hilft bei …?"): use search_passages, answer FROM the returned fulltext passages, and cite each source (name + source_url/page). Phrase the question as a topical statement naming subject and target group ("Friedenserziehung in der Grundschule: Einstieg mit Kindern"), not the user's literal "Wie kann ich …?" — ranking is hybrid keyword+vector. A passage with a snippet but no text has no fulltext yet; say so. No source restriction from the user? Just pass the relevant content kinds (e.g. kinds:[30142]). Optionally follow up with search_content to offer browsable materials. - DISCOVERY intent — the user wants materials to browse ("finde/suche/empfiehl Materialien zu X"): use search_content and present the items as links. - AMBIGUOUS or very broad: ask ONE short clarifying question ("Möchtest du eine begründete Antwort aus den Materialien, oder eine Liste von Materialien zum Stöbern?") — and mention scoping options the user may not know exist: restricting to a publisher or author, a specific relay/corpus (list_relays), or a time window. When the user restricts the sources ("nur Content von X", "only from X"), route the restriction into scope parameters, never into the question/query text: a metadata publisher (most organisations: resolve_publisher(name) finds the exact spelling) goes into search as a quoted field filter, e.g. search:'publisher.name:"LEHRE LADEN"'; a Nostr signing account (resolve_author(name) → pubkey) goes into authors. Publisher names must be quoted inside the search value — an unquoted multi-word name silently matches nothing. Authorship has two distinct layers — never conflate them: eventAuthor is the Nostr signer/uploader (often an aggregator), while creator/publisher (on resources) are who actually made and published the material. "Who published this?" is answered by publisher, never by eventAuthor. Searches run against the default relay set. list_relays may advertise extraRelays — additional relays holding different corpora (e.g. a broader aggregation) that are only queried when you pass them via the relays parameter of search_content, search_resources, or get_resource. When a search on the defaults comes up short, or the user asks for a specific relay's holdings, check list_relays and re-search with relays set; fetch follow-up details with the same relays value the search used. The server also browses controlled vocabularies (browse_*, skos_* tools) and, for authenticated clients, signs and publishes new metadata (signer_*, create_and_publish_*).