Category id 7a2651313875c553 (derived from Scout's category name (no id in Scout's data)). 17 alternatives, shown in Scout's order.
Scout's category view
Search APIs for agent web research (serper.dev deep dive, 2026-07-23). Tiered pick: Brave+Tavily primary (independent index + LLM-native, clean legal, MCP), DataForSEO/serper for bulk Google, SearXNG self-host fallback, Jina utility. Bing retired 2025-08-11; Google CSE closing.
Scout's judgement, from alternatives.json (2026-11-18). It is not a gaents rating.
Our own measured runs
SERP feasibility on Apify (2026-10-20) (internal: reports/market/serp-feasibility/RESULTS.md (not published))
Alternatives
serper.dev
commercialpage only: no Run-it in this category
Google SERP scraping-as-a-service. Endpoints: search, images, news, maps, places, videos, shopping, scholar, patents, autocomplete, scrape. Structured JSON (organic, knowledge graph, answer boxes, PAA, related). Widely used in the LangChain/agent ecosystem. (Scout's description)
No published SLA; '1-2s' is a claim not measured. 6-month credit-expiry trap (user reported losing $375 balance). LEGAL: terms disclaim all warranties + mutual indemnification — YOU defend serper for claims from your use. Zero legal cover on scraping Google's SERPs. Cheapest raw-Google but never load-bearing without a fallback.
Lock-in risk
low
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T22:08Z cloud refresh: [2026-10-19 Scout T2 price/licence check: Prices same as the stored secondary-sourced figures; now confirmed from serper.dev home page pricing section.]] [2026-10-20T01:33Z cloud refresh: type proprietary -> commercial (no recurring free tier: 2,500 free queries at signup (one-off), then prepaid packs; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from terms; api_or_webhooks from home page. self_hostable left out (no page says hosted-only). data_export n/a (stateless search API).]
Brave Search API
freemiumpage only: no Run-it in this category
The only non-Google INDEPENDENT index in the field — own index, 30B+ pages, 100M+ updates/day (not a scraper repackaging Google/Bing). Endpoints: web, LLM-context, answers, image, video, news, suggest, spellcheck. Reference index behind Anthropic's brave-search MCP server. (Scout's description)
CONFIRMED 2026-10-19: Prepaid credits, no subscription. Search $5 per 1,000 requests, includes $5 in free credits every month, 50 queries per second. Answers $4 per 1,000 requests + $5 per million input/output tokens, includes $5 in free credits every month, 2 queries per second. Headline: 'Up to $20 in free calls every month'. Enterprise: custom terms, full-funnel Zero Data Retention. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
Index is very good but NOT identical to Google — for exact SERP-position work it differs; for agent research it's more than adequate. $5/1k pricier than serper/DataForSEO but far cheaper than SerpApi. The cleanest legal footing: owns the index → no scraping-liability for us at all.
Lock-in risk
low
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
[2026-10-19T22:08Z cloud refresh: 2026-10-19 freshness check: per-request prices unchanged. Billing model now stated as prepaid credits with no subscription, and the headline free allowance is 'Up to $20 in free calls every month' (the page does not say how the $20 splits across products; per-plan blocks each show $5). Answers capacity is 2 QPS (stored text did not give it). Search still 50 QPS.] [2026-10-20T01:33Z cloud refresh: type proprietary -> freemium (proprietary; pricing has free tier + paid; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class and api_or_webhooks from Brave Search API terms; self_hostable left out; data_export n/a (search API).] [2026-10-23T13:15Z cloud refresh: Fact fields pass 2 (2026-10-20): self_hostable false: hosted API over Brave's own index (endpoint https://api.search.brave.com/res/v1/; the brave.com/search/api page, which also carries the plans, offers no self-hosted option, agent-checked). data_export n/a (search API).]
Tavily
freemiumpage only: no Run-it in this category
Search API built FOR agents: returns LLM-optimized output — cleaned, ranked, deduped snippets + optional synthesized answer + raw content in one call. Removes the scrape→clean→chunk step. Official Tavily MCP server. (Scout's description)
Free 1,000 API credits/mo (no card). PAYG $0.008/credit (~$8/1k advanced). Project + Enterprise tiers. Basic=1 credit, advanced=2. (verified primary) as of: no date recorded for this field
Scout, alternatives.json v3.72026-11-09Scout last_verified; no date recorded for this field itself
Proprietary response shape (answer+snippets) is agent-friendly but not a drop-in for Google JSON; swap cost moderate. Aggregates + processes (not raw Google-SERP repackaging) → sidesteps the ToS grey zone. Best fit for how our agents actually consume search.
Lock-in risk
low
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
[2026-10-19T22:08Z cloud refresh: Q4-2026 review: Tavily announced on 2026-02-10 that it is joining Nebius. It says the API stays the same and it operates as before. Pricing re-checked, unchanged. Verdict unchanged.] [2026-10-20T01:33Z cloud refresh: type proprietary -> freemium (proprietary; pricing has free tier + paid; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from Tavily terms; self_hostable false: hosted API (api.tavily.com) with no self-hosted edition (API reference page); api_or_webhooks from API reference; data_export n/a.]
Jina Reader / Search (r.jina.ai, s.jina.ai)
freemiumpage only: no Run-it in this category
r.jina.ai converts URL→clean markdown; s.jina.ai returns top-5 results as LLM-ready text. URL-prefix API — trivially droppable. (Scout's description)
CONFIRMED 2026-11-09: Reader (r.jina.ai) free without API key at 20 RPM, 500 RPM with free or paid key, 5,000 RPM premium. Search (s.jina.ai) blocked without key; 100 RPM with free or paid key, 1,000 RPM premium; each search costs a fixed token amount starting at 10,000 tokens. Token-billed with key; every new API key gets 10M free tokens. as of 2026-11-09
Scout, alternatives.json v3.72026-11-09date written in Scout's pricing text
Reader ~8s latency. But near-zero-cost URL→markdown is genuinely useful glue for our fetch/clean step regardless of which search API we pick. s.jina.ai is a fine free fallback search.
Lock-in risk
low
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
[2026-10-19T22:08Z cloud refresh: [2026-10-19 Scout T2 price/licence check: Prices and limits unchanged: r.jina.ai 20 RPM without key, 500 RPM with free key; s.jina.ai needs a key, 100 RPM free; every new key gets 10M free tokens. Ownership note: the official Reader page now routes contracts, invoices and on-prem licences to Elastic, and says Elastic has sold the Jina commercial licence as its own SKU since August 10, 2026. Some reader models (ReaderLM-v2, jina-vlm) are CC-BY-NC 4.0.]] [2026-10-19T22:08Z cloud refresh: Q4-2026 review: ownership. jina.ai pages now carry 'Elastic (c) 2026', and Elastic has sold the Jina On-Prem commercial licence 'since August 10, 2026'. Self-serve API keys still bought via Stripe on jina.ai. Reader models ReaderLM-v2 and jina-vlm are CC-BY-NC 4.0 (quoted as fact). Verdict unchanged. Latest jina.ai news post 2026-10-06. Pricing for this entry is carried by patches/02-price-refresh.json (fuller text, same facts); apply 02 before 03.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class proprietary (hosted API; on-prem only under a commercial licence sold by Elastic); self_hostable true only via the paid Jina On-Prem containers; api_or_webhooks from Reader page; data_export n/a (stateless API).]
DataForSEO
commercialpage only: no Run-it in this category
Enterprise-grade Google SERP + SEO data API. Cheapest $/query in the field for batch. Standard queue is async (task-post/task-get); Live endpoint real-time at 3× price. (Scout's description)
$0.0006/request standard queue (~$0.60/1k), $0.002/request live (~$2/1k). $50 min deposit. Billing base depth 10 results. (verified help-center; per-req figures partly secondary) as of: no date recorded for this field
Scout, alternatives.json v3.72026-10-23T13:15ZScout patch_history.at; no date recorded for this field itself
Async standard queue not ideal for interactive agent latency; Live is real-time but pricier. Verbose task-queue API (stickier than a plain GET). Best for batch/corpus enrichment (tweet-corpus, bulk fact-checks). Same scraping legal posture (no shield).
Lock-in risk
medium
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-20T01:33Z cloud refresh: type paid -> commercial (proprietary; pricing paid; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: api_or_webhooks from DataForSEO v3 docs; licence_class left out (terms page has no licence wording); self_hostable left out; data_export n/a (data API).] [2026-10-23T13:15Z cloud refresh: Fact fields pass 2 (2026-10-20): self_hostable false: hosted API (docs base address https://api.dataforseo.com/; dataforseo.com/pricing is pay-as-you-go with no self-hosted option, agent-checked). licence_class left out (terms have no licence wording). data_export n/a (data API).]
SerpApi
freemiumpage only: no Run-it in this category
The incumbent Google SERP API — broadest engine coverage, publishes performance/uptime, de-facto JSON schema standard. Headline differentiator: 'US Legal Shield' (claims to assume legal liability for scraping). (Scout's description)
CONFIRMED 2026-10-19: Free $0 (250 searches/mo); Starter $25/mo (1,000); Developer $75/mo (5,000); Production $150/mo (15,000); Big Data $275/mo (30,000); month-to-month. Only successful searches count. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
~10-15× serper's price for identical Google data. The ONLY reason to pay the premium is the legal shield — and its pricing page NAMES the feature but doesn't spell out the indemnity; VERIFY THE ACTUAL MSA before relying on it as real cover.
Lock-in risk
low
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T23:00Z cloud refresh: 2026-10-19 price refresh: unchanged. type 'paid' -> 'freemium' (free plan).] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: api_or_webhooks from Search API docs; self_hostable false: hosted API only (endpoint on serpapi.com; no self-hosted edition offered). licence_class left out: the legal page has no licence-grant clause to quote. data_export n/a (stateless search API).]
SearchApi.io
commercialpage only: no Run-it in this category
Multi-engine SERP API (50+ engines) with a published 99.9% SLA and pay-per-success billing. (Scout's description)
CONFIRMED 2026-10-19: 100 free requests; Developer $40/month (10,000 searches, $4/1k) stepping down to Octo 5M $5,000/month ($1/1k). Billed only for successful (HTTP 200) searches; 99.9% SLA; up to $2M legal-protection coverage. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
The 'reliable middle' — cleaner/more reliable than serper, pricier than DataForSEO. Same scraping legal posture (no shield). Reach for it if serper's no-SLA bites and DataForSEO's async doesn't fit.
Lock-in risk
low
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T23:00Z cloud refresh: 2026-10-19: type 'paid' is not an allowed value; set to commercial.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: api_or_webhooks from Google engine docs; self_hostable false: hosted API only (endpoint on searchapi.io; no self-hosted edition offered). licence_class left out: terms page has no licence-grant clause to quote. data_export n/a (search API).]
Exa
commercialpage only: no Run-it in this category
Neural/embeddings search API for AI (formerly Metaphor). Best for semantic 'find things LIKE this' queries, discovery, findSimilar, research fan-out — where you don't have exact keywords. (Scout's description)
Weaker/pricier than a SERP API for cheap factual lookups. A COMPLEMENT to a keyword API for discovery-style research, not a general replacement. Neural results are Exa-specific (no equivalent elsewhere → med lock-in). Also listed in the Search category.
Lock-in risk
medium
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout tested it (proven, 2026-07-25): Same as the Search-category entry — kept in sync
Scout's notes
Verdict evaluate -> adopt 2026-07-25 on MEASURED evidence, not marketing: ran it on 10 real UK SMEs for Pixel. $0.007/pass, $0.063/company for 67 fields, 40.9 real fields/company, 1.4-2.2s/query. Per-field grounding citations + confidence are the standout feature. TWO MANDATORY USAGE RULES: (1) key queries on a registry identifier (CH number + postcode) — a bare company name makes it confabulate a confident WRONG entity; (2) outputSchema caps at 10 properties counted RECURSIVELY, so wide sweeps need multiple passes. CAVEAT: answers 'don't know' as PROSE not null — 29% of apparently-populated fields were sentences meaning nothing until filtered. [2026-10-19T22:08Z cloud refresh: Changed: the page now lists a Lite Search lane at $4/1k ($0.004/search), below the stored $0.007/search (which matches Auto Search, unchanged). Plan names are now Starter / Developer / Enterprise, plus a $10 onboarding bonus on the free plan. Page banner: 'Introducing Exa 3.0: Our largest ever improvement to search quality'. Auto price unchanged, so two_year_cost_gbp not touched.] [2026-10-20T01:33Z cloud refresh: type paid -> commercial (proprietary; pricing paid; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from Exa terms PDF; api_or_webhooks from API docs. self_hostable left out. data_export n/a (stateless search API).] [2026-10-23T13:15Z cloud refresh: Fact fields pass 2 (2026-10-20): self_hostable false: Exa is a hosted search API over Exa's own index; the Enterprise page (the page that lists plan options: ZDR, SLAs, SSO, custom MSA) offers no self-hosted or on-prem edition (agent-checked). data_export n/a (stateless search API).]
SearXNG
open-sourcepage only: no Run-it in this category
Self-hosted METASEARCH aggregator — holds no index, queries ~200 engines (Google/Bing/Brave/DDG/Startpage/Mojeek/Qwant/…) in parallel over plain async HTTP (httpx[http2] + httpx-socks), de-dupes/re-ranks, strips tracking. Actively-maintained successor to searx. JSON API via &format=json (off by default — enable formats:[json] + token-gate). Proxies (HTTP + SOCKS/Tor) are first-class. (Scout's description)
Fetches via httpx, NOT a browser — so it has no browser fingerprint and Camoufox/patchright CANNOT help its blocking (they're the scrape-layer cure; SearXNG's is proxies + limiter + engine rotation). At volume, upstream engines (Google worst) block on datacenter-IP reputation + rate — a bare single-VPS instance is NOT dependable for high-frequency Google queries without a residential-proxy pool. Genuinely good as a free tier-0 fallback + for non-Google engines. Engine-breakage upgrade treadmill is the real maintenance cost.
Lock-in risk
low
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
[2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (AGPL-3.0); self_hostable from README install guide; api_or_webhooks from search API docs; data_export n/a (metasearch).]
Firecrawl /search
freemiumpage only: no Run-it in this category
Primarily a scrape/crawl product with search bolted on. Official MCP server. (Scout's description)
Free 1,000 credits/mo. Hobby $16 · Standard $83/100k. Search = 2 credits/10 results. MCP server. (verified primary) as of: no date recorded for this field
Scout, alternatives.json v3.72026-10-23T13:15ZScout patch_history.at; no date recorded for this field itself
Reach for it only if we need crawl+scrape+search unified; not the cheapest pure-search. (FireCrawl already in Web Scraping / Crawling category.)
Lock-in risk
low
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: Correction: stored license "Proprietary (SDK open)" is wrong; the Firecrawl README says the project is AGPL-3.0 with MIT SDKs, and the cloud adds features (stored type left as is). licence_class copyleft for the open core. api_or_webhooks from /search docs. self_hostable left out: a self-host guide exists for Firecrawl, but I did not confirm /search runs self-hosted. data_export n/a (pure API).]
ScrapingDog
freemiumpage only: no Run-it in this category
serper-class Google SERP scraper + general-purpose scraper. (Scout's description)
CONFIRMED 2026-10-19: Free $0 forever (100 credits); Lite $40/month (200,000 credits); Standard $90 (1M); Pro $200 (3M); higher tiers to Business Pro $1,500 (29M). Google SERP Light API 5 credits/request (Lite = 40K requests, $1/1K); Google SERP Advance API 10 credits/request. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
CONFIRMED 2026-10-19: Free 50 searches/mo; Small $49.99/mo (25,000 searches); Medium $149.99/mo (100,000); Large $299.99/mo (250,000); Premium $499.99/mo (500,000); Enterprise $899/mo (1,000,000). Cost per 1,000 falls from $2.00 to $0.90. Only successful requests billed. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
serper-class; no distinguishing advantage. Volume figures secondary.
Lock-in risk
low
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T23:00Z cloud refresh: 2026-10-19 price refresh: entry plan $49.99 now gives 25,000 searches (stored said 5k). Now first-party sourced. type 'paid' -> 'freemium'.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: api_or_webhooks from home page (SERP API with API key). licence_class and self_hostable left out: terms URLs redirect to the home page; no cloud-only statement. data_export n/a (search API, holds no user data).]
Serply
commercialpage only: no Run-it in this category
Google SERP scraper, prepaid credits. (Scout's description)
CONFIRMED 2026-10-19: 2,500 free credits (30 days, no card); prepaid packs from 2,500 credits for $5 up to 10,000,000 credits for $7,500; rate $2.00/1k (Micro), $1.50 (Starter), $0.98 (Growth), $0.75 (Pro and above). Credits never expire; cached responses free. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
serper-class; no advantage. Confirm pricing on live quote.
Lock-in risk
low
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T23:00Z cloud refresh: 2026-10-19: old '10k for $15 -> 2.5M for $1,249' replaced by vendor pricing page figures. type 'paid' is not an allowed value; set to commercial.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class proprietary from Serply Terms (limited licence to use and access the Site; IP owned by Company); api_or_webhooks from home page (one REST API for Google search and more). data_export n/a (search API). self_hostable left out.]
ScrapingBee
commercialpage only: no Run-it in this category
General web scraper with a SERP mode. (Scout's description)
~$2/1k SERP; a general scraper first. Reach for it only if we also need general JS-scraping.
Lock-in risk
low
Integration hours (estimate)
3
Credited to Scout (2026-11-18T22:56Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T23:00Z cloud refresh: 2026-10-19: stored '$40/mo ≈ 20k SERP' plan no longer exists; cheapest is Hobby $19/mo for 7,500 searches.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from terms (right to use the API; no reverse engineering); self_hostable false: hosted API only (endpoint on app.scrapingbee.com; no self-hosted edition offered); api_or_webhooks from docs. data_export n/a (stateless scraping API).] [2026-11-18T22:56Z cloud refresh: Apify-alternatives review 11-18: plans Hobby $19/75k credits, Freelance $49/250k, Startup $99/1M; JS render 5 credits, premium proxy 10 (25 with JS). JS page on Freelance = $0.00098 (premium+JS $0.0049) vs our site-render ~$0.0020/page. Terms §6 (Conditions of Use): users shall not 'Use the Services or any element thereof in a commercial manner.'; IP clause bars 'making the API available to a third party'; §16 'User owns all rights, title, and interest ... in and to its User Data.']
Google Custom Search JSON API
commercialpage only: no Run-it in this category
Official Google search JSON API — but site-restricted and never a general web-search option. (Scout's description)
CONFIRMED 2026-10-19: existing customers only: 100 queries/day free; $5 per 1,000 queries, up to 10k/day. Closed to new customers; service discontinued 2027-01-01. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
Closing to new customers; site-restricted; 100/day free. Dead for us. (verified Google overview 2026-07-23)
Lock-in risk
medium
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
[2026-10-19T23:00Z cloud refresh: 2026-10-19: closure and 2027-01-01 discontinuation re-confirmed. Verdict stays skip.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from Google APIs Terms (no reverse engineering or source extraction); api_or_webhooks from overview; self_hostable false: hosted Google API only. data_export n/a (search API). Official notice re-read: closed to new customers, discontinued 2027-01-01 (verdict already skip; unchanged).]
Bing Search API
commercialpage only: no Run-it in this category
Was Microsoft's web search API. RETIRED hard cutoff 2025-08-11. Replacement is Grounding with Bing Search inside Azure AI Agents (different, pricier, LLM-coupled). (Scout's description)
CONFIRMED 2026-10-19: Search $0.75 per 1,000 successful requests; Fetch $0.75 per 1,000 successful requests. QPS limit per account: Search 100, Fetch 25 (separate quotas; Fetch also has a per-domain limit). Server-side tool calls via DeepSeek Harness, Claude Code or Codex are free until per-request billing starts 00:00 Beijing time on November 1, 2026. as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
One triage pass; re-verify stars, licence and price before acting.
Limits for us
See research/tweet-triage-2026-10-18-scout.md (section for DeepSeek Search & Fetch API).
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Merged from research/tweet-triage-2026-10-18-scout.md by merge_triage_table.py. [MEASURED 2026-10-18: LinkedIn finding 3/200 right-first vs Tavily 166, Serper 152 (index has few /in/ profiles); Fetch reached 33/50 vs site-render 42, but returns trimmed main content (no footers): 0/11 company numbers. Search = skip for person finding; Fetch = evaluate for article text only. research/tweet-triage-2026-10-18-scout.md 6b] [2026-10-19T22:08Z cloud refresh: 2026-10-19 freshness check: price unchanged; rate limit changed. Stored '5 QPS' is out of date: docs now say Search 100 QPS and Fetch 25 QPS, counted separately. Docs: 'These calls are currently free of charge; per-request billing starts at 00:00 Beijing time on November 1, 2026.' (applies to server-side tool calls).] [2026-10-20T01:33Z cloud refresh: type None -> commercial (proprietary; pricing paid; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from DeepSeek Open Platform terms; self_hostable false: hosted API on api.deepseek.com, no self-hosted edition (search guide); api_or_webhooks from the same page; data_export n/a.]