X / Twitter Search
Search X (Twitter) with full advanced-search syntax (operators like from:, since:, until:, min_faves: work inline in the query) and get structured tweets with text, author, engagement, and cursor pagination. Limit is a per-page maximum; native pages contain approximately 20 tweets unless requireSinglePage selects a bulk lane.
Price: $0.75 per 1,000 requests (flat per request - same cost regardless of results returned).
Routing: 2 independent lanes - a failed attempt fails over to the next lane automatically, and the same normalized schema comes back either way.
Authorizations
Your AnyAPI key as a Bearer token.
Headers
Optional wallet idempotency key, scoped to this customer for 24 hours. When the gateway honors the key, this synchronous in-process execution can continue after the caller disconnects, bounded by its execution deadline. A completed replayable result charges normally exactly once and can be replayed without another provider run or charge. A pending duplicate returns 409 idempotency_in_progress; reuse with different request semantics returns 409 idempotency_conflict.
1 - 255Query Parameters
Optional. Comma-separated keys (dotted paths like author.name descend into nested objects) to keep on each result item. Keys are matched relative to each result item after the data/items envelope is unwrapped, not against the top-level response envelope, so use jq to reshape the whole envelope. Shrinks the response without changing cost.
Optional. Cap the number of result rows returned; a _truncated note reports how many were withheld so you can page via the API's own limit. Does not change cost.
x >= 0Optional. Return only a structural outline (top-level keys, item counts, and per-field byte sizes) instead of the full data. Does not change cost.
Optional. A jq expression applied to the result envelope; its output replaces output (multiple outputs collect into an array). Reshape freely, e.g. jq=.data | {title, description, md: .markdown[:3500]}. Runs sandboxed with a 250ms / 2MB budget; on failure the full result is returned with a jqError. Does not change cost.
Body
Search query using X (Twitter) advanced-search syntax. IMPORTANT: bare terms are ANDed - a tweet must contain EVERY word, so a list of loosely related keywords matches nothing; use one short phrase or OR between alternatives (e.g. 'anyapi OR getanyapi'). You can embed X advanced-search operators directly in the query to filter results: from:username and to:username (author or recipient), since:YYYY-MM-DD and until:YYYY-MM-DD (date range), min_faves:N, min_retweets:N, min_replies:N (engagement floors), "exact phrase", -term to exclude, filter:media and filter:links and -filter:replies (content filters), lang:en, near:city, and geocode:lat,long,radius. Examples: 'from:OpenAI', 'AI agents min_faves:500 -filter:replies', 'nvidia since:2026-01-01 until:2026-03-01'. A query with no matches returns an empty items array; prefer the fewest words that identify the topic.
Opaque pagination cursor from a previous response's nextCursor. Omit for the first page; pass it to fetch the next page of search results.
Optional ISO 639-1 language code to restrict tweets to (e.g. en).
Per-page maximum number of results to return (1-50, default 20). A provider may return a smaller native page; follow nextCursor for more.
1 <= x <= 50Result ranking: 'Latest', 'Top', 'Photos', or 'Videos' (e.g. Latest).
Set true to get up to limit results in one response instead of provider-native pages, served by a bulk provider when needed.
Response
Normalized result.
USD charged on the original run. On a replay this value is echoed for parity; the replay itself is free.
Normalized output, or null when the replay payload was not retained.
Always "AnyAPI".
True when this response replays the durable result of an earlier run without billing or upstream execution.
Number of result rows returned. For per-result SKUs the per-item cost is charged against this count; for input-priced SKUs the charge is per submitted input, independent of this count.
Present only when a jq expression failed; output then carries the full unshaped result and this explains why the reshape did not apply.
Opaque handle to the full unshaped result, cached ~15 min. Re-shape it for free (fields/max_items/summary/jq) via GET /v1/results/{id}, no re-billing. Absent when the result was too large to cache.