The corpus, over plain HTTP

Reading needs no key

The corpus is public on the website, signed out and crawlable, so gating the same rows here would buy nothing and would cut this API out of the way most people find it. Only the endpoints we are prepared to keep working are documented: an endpoint that exists is not the same as an endpoint that is a promise.

What a key buys

Volume, and one field. A key raises the rate limit from 60 calls an hour to 600, and a post's mirrored media ships as a real URL rather than as null with mediaLocked set. Send it as Authorization: Bearer <key>.

True of every endpoint here

• Every list endpoint takes dataset and means it. A corpus is a subject, and a search across the whole library answers a question nobody asked.
• Every post carries its outlier multiple beside its engagement. Rank on the multiple: raw reactions only find accounts with big followings, and a 200-reaction post from somebody who averages 30 is the one worth stealing from.
• A repost is attributed to whoever wrote it, not to the account that reshared it. Reposts are hidden by default, and reposts=all is how you ask for them.

What is documented

8 endpoints across the corpus, search, the verticals, people and the MCP server, each with copy-paste examples in cURL, JavaScript, Python, Go, Ruby and PHP.