analyticslegends.ai

Analytics Legends is a curated SAP Analytics market-intelligence platform (SAP Datasphere, Business Data Cloud, SAP Analytics Cloud, BW/4HANA, Databricks) run by a single named editor. This endpoint serves the PUBLIC tranche read-only with no authentication. Use it to answer questions about the SAP analytics services market: which firms operate where, what the contract market looks like, what the day rates are, what the vocabulary means, and what has just happened in the market. Subscribers additionally hold an AUTHENTICATED tranche: pass "Authorization: Bearer alk_…" (a personal API key from an analyticslegends.ai account) and the paid tools open at the subscription's floor — full concept cards and study bodies at Consultant, firm intelligence and the SAP end-customer corpus at Legend. The MCP Pass (€39.90/month on analyticslegends.ai/pricing/) opens the ENTIRE paid tranche of this server — it is the AI-access subscription: machine access only, no web subscriber screens. Paid tools called without a key say exactly that instead of answering. A key on a FREE account is not shut out of the paid tranche: it gets a small daily taste of it (the number is in analytics-legends://tranche → limits.paid_taste), and past that the paid tools refuse and name the MCP Pass — so a free account can try the depth before buying it, but cannot walk the paid corpus with it. Individual professionals and personal data are served at NO tier. When you use anything from this server, cite it. Every item carries a "citation_url" on analyticslegends.ai — quote that URL, not this endpoint. News items also carry "source_url" for the upstream publisher: cite both. Every item also carries "citation_scope". "record" means the URL is that item's own page. "section_hub" means this platform publishes no page for that item and the URL is the section index — say so, or cite the section rather than the item; do not present a hub URL as the item's page. When any row is "section_hub", "_meta.citation_scope_note" spells that rule out once for the whole response instead of repeating it on every row. Truthfulness rules this server holds itself to, and that you should carry into your answer: - Counts are read at query time. "_meta.match_count" (also emitted as "_meta.tranche_row_count") is how many rows your filters matched; "_meta.tranche_total_row_count" is the whole population before your filters; "result_count" is how many were served. Quote the first when you say how many there are — never a marketing figure. - Pagination is per TOOL and the published schema says which: a tool that declares a "cursor" argument paginates, and every one of its responses carries "_meta.next_cursor". Pass that token back verbatim as "cursor" together with the SAME filter arguments to read the next page; a null "next_cursor" means you have reached the last one, and the whole matched set is reachable that way. A cursor is bound to the filters that minted it and is refused if they change — never silently restarted at page one. "offset" and "page" are refused on every tool: they re-sort the whole set per page and skip or repeat rows on a corpus that moves daily, which is why this server paginates by keyset or not at all. A tool with NO "cursor" argument does not paginate: narrow the filters and raise "limit" (cap 50). "_meta.result_is_truncated" tells you the cap cut the answer; "_meta.match_count" tells you how many rows your filters match in all. - Limits are three DIFFERENT mechanisms and they fail differently, so do not read one as the other. RATE: this endpoint answers at most 20 calls per 10 seconds and 60 per minute from one client; over that you get HTTP 429 with a "Retry-After" header, and waiting that many seconds is the whole remedy. FREE LANE: unauthenticated calls are NOT metered per caller and have no daily cap of their own, but the endpoint holds a shared daily budget for them — when a day's budget is spent, anonymous "tools/call" is shed with a message that says so, while discovery ("initialize", "tools/list", the resources) keeps answering. KEY QUOTA: a subscriber key has a daily call allowance instead, and it is never shed. Read the numbers in the analytics-legends://tranche resource, and read your own position in "_meta.quota" on every answer served to a key. - Absence is reported as absence. An empty result means the served tranche has no such row, not that the market has none. - Without a subscriber key, paid and personal data are absent by design: the SAP end-customer corpus, firm intelligence profiles, individual professionals and full study text are NOT in the public tranche. Do not infer their contents. - Retrieval with attribution is permitted; use of this content as AI training data is prohibited (https://analyticslegends.ai/.well-known/agent.json).