The High Wall of the Login-Free Garden: Xiaohongshu Verdict of Zero Apertures, Three Login-Free Pipelines of WeChat Official Accounts, and the Final Accounting of the Free-Scraping Map

After the map of X was finished, the same methodology of rounds-plus-field-tests was marched toward two new fronts — and returned holding a mirror: the openness of X and the closedness of Xiaohongshu are the two poles of the modern anti-crawling ecology.

I. Xiaohongshu: The Verdict of Zero Apertures

Ten rounds plus twenty plus supplements settle it cold and hard: there is no login-free, key-free data aperture. Every API on edith.xiaohongshu.com returns 406 or no-login-info unauthenticated; there is not even an og meta tag (not even the metadata aperture X offers); public RSSHub instances carry no route; the x-s/x-t signature scheme drifts with versions and no standalone signing library lives; short links resolve but land on the same signature wall; the CDN runs hotlink protection. The open-source consensus route is MediaCrawler (69k stars) — compute the signature inside a live browser, bypassing reverse engineering, at the cost of scanning to log in once with your own account. The commercial platforms (Newrank/Qian-gua) offer free tiers of leaderboards only. The difference of risk dimensions is the crux: X polices by IP (change IP, solved), Xiaohongshu by account-plus-device-signature (a real login state is mandatory) — which is why X could be farmed through GitHub Actions while Xiaohongshu cannot even sustain a key-free public tool.

II. WeChat Official Accounts: An Unexpected Harvest of Three Pipelines

Xiaohongshu’s opposite, the WeChat official account, yielded three field-tested login-free routes. Sogou WeChat search (weixin.sogou.com) returns real results without login, twenty per page, stable pagination, six rapid calls without throttling — the round’s biggest find. Direct article links (mp.weixin.qq.com/s/) pour out full HTML with a URL; a fake ID correctly returns the content-not-found page rather than 403 — no request-level interception. And the /link relay page of Sogou, its JS fragments reassembled, yields the real signed article address — the search→resolve→fulltext pipeline is open end to end, written into collect_wechat.py and merged into the tool (the trial run captured real full texts from outlets like QbitAI; dedup by stable title hash). There is also wewe-rss (9.6k stars) as a self-hosted bridge: WeRead-based credentials, one scan-to-login subscribes any account. The structural resemblance to X is startling: Sogou ≈ Nitter search RSS, article links ≈ the syndication CDN, wewe-rss ≈ a self-built Nitter.

III. Eighteen More Rounds on X: From Gleaning to the Ceiling

The map of X also expanded before it converged: the three verifications dug out the syndication CDN single-tweet endpoint (token derivable from the tweet ID by mathematics, login-free and key-free, the steadiest of the three detail sources); five rounds dug out the trends24 trend site (unauthenticated global and per-region hot tags) and Shitter (the live Nitter fork’s 21-instance table, landing dual-source instance discovery); ten rounds dug out Archive.org’s Twitter Stream Grab (Archive Team’s rescued 2006-2023 full tweet stream, 130+ monthly items, the honest road to free historical data). Two negative findings matter equally: the guest token activates but GraphQL answers tourists with 404/422 — the login-free channel was sealed by X in 2026; and the x-client-transaction-id library proves the anti-crawling is still climbing. The final map settles at six classes; 500-5,000 items a day per IP, past ten thousand with Actions IP rotation.

IV. Where Cartography Ends

Fifty-odd rounds, a dozen search engines, forty-plus endpoints field-tested — the survey draws its own border: the lightweight free interfaces are exactly RSS mirrors, embedded JSON, trend sites, and the historical archive, four classes; beyond that lies the land of accounts and payment. The shelf life of channels is days to weeks (vxTwitter went from 200 to 403 within hours), but the replacement mechanism does not die. The true product of this survey is not a channel list but a tool that discovers, verifies and replaces its own channels — maps expire; the craft of surveying does not.