> ## Documentation Index > Fetch the complete documentation index at: https://docs.xquik.com/llms.txt > Use this file to discover all available pages before exploring further. # Xquik changelog: Twitter API & scraper updates > Track Xquik REST, MCP, SDK, tweet search, follower export, monitor, webhook, extraction, account action, and contract changes. Newest changes come first.
For the complete documentation index, see llms.txt.
Search tweets, API keys, signed webhooks, user profiles, REST, MCP, authentication, billing, and writes all appear in this changelog. **New** * The 6 analysis endpoints take `baseline`: the rows of an earlier answer. Each result's `monitor` says whether the post is new, changed or unchanged. The comparison costs nothing extra * Score answers, such as `intensity` or `conviction`, can have decimals between levels * 6 AI analysis endpoints read X posts, searches or your own texts. Each analyzed post costs 2 credits * `POST /x/analysis/sentiment` answers each post's attitude, its intensity & its sarcasm probability * `POST /x/analysis/brand` answers each post's relevance, attitude & customer relationship for your brand * `POST /x/analysis/news` answers each post's format, source attribution & relevance to your topic * `POST /x/analysis/market-signals` answers each post's stance on an asset & counts bullish & bearish posts per cashtag * `POST /x/analysis/viral-score` rates how viral a post or draft reads, from 0 to 100, with a verdict * `POST /x/analysis/classify` answers up to 8 questions you write: categories, scores or yes/no * A search can name an account in `username`, a List in `listId`, or a post in `quotesOf` or `repliesTo`, between `sinceTime` & `untilTime` * A post without an analysis costs nothing. MCP lists all 6, and guest keys can call them * `GET /x/articles/{tweetId}` adds `bodyMarkdown`: the Article body as Markdown * It keeps headings, lists, bold, italic, links & images, without HTML. It costs nothing extra * Search rows add `textHighlights`: where the row's `text` matches the search * Each range has `startIndex` & `endIndex`, counted in code points of `text`. It costs nothing extra * `GET /x/users/batch/tweets` returns the newest posts of up to 20 users in 1 call * Name the users by `ids` or `usernames`. It costs 1 credit per post * A user it could not read is named in `unprocessed_ids` & costs nothing * MCP adds it as `getBatchUserTweets`, and guest keys can call it * `GET /x/communities/find` finds X Communities by keyword. It costs 1 credit per Community * `GET /x/communities/popular` lists the most popular Communities of a topic or subtopic * Send the topic by `topicId` or by name as `topic`. It costs 1 credit per Community * `GET /x/communities/topics` lists every Community topic & subtopic, 1 credit per topic * `GET /x/communities/suggested` lists the Communities X suggests to join, 1 credit each * `GET /x/communities/{id}/media` returns a Community's Media tab, 1 credit per post * `GET /x/communities/{id}/tweets` takes `queryType=Top` for the Community's Top posts * Community responses add `banner`, with its size, main colors & focus area * MCP adds `findCommunities`, `getPopularCommunities`, `getCommunityTopics`, `getSuggestedCommunities` & `getCommunityMedia`, and guest keys can call them * Community find & popular answer a free 400 `invalid_cursor` when X refuses a cursor * `GET /x/tweets/{id}/retweeters/check` says whether 1 account reposted a post * It reads the account's newest posts. Each check costs 1 credit * MCP adds it as `getTweetRepostCheck`, and guest keys can call it * `GET /x/users/{id}/affiliates` returns the accounts an organization lists as its affiliates, in X's order * Each row is a full profile. It costs 1 credit per result * MCP adds it as `getUserAffiliates`, and guest keys can call it * `GET /x/users/{id}/subscriptions` returns the creators a user subscribes to, in X's order * Each row is a full profile. It costs 1 credit per result * MCP adds it as `getUserSubscriptions`, and guest keys can call it * `GET /x/users/{id}/follower-ids` and `GET /x/users/{id}/following-ids` return user IDs only * A page holds up to 5,000 IDs, newest first. Each ID costs 1 credit * MCP adds them as `getUserFollowerIds` and `getUserFollowingIds`, and guest keys can call them * `GET /x/tweets/{id}/hidden-replies` returns the replies a post's author hid, oldest first * It lists replies to hidden replies too. It costs 1 credit per reply * MCP adds it as `getTweetHiddenReplies`, and guest keys can call it * `GET /x/tweets/{id}/translation` returns X's own translation of a tweet into the language you pick * It returns the detected source language and the links of the translated text. It costs 1 credit per call * MCP adds it as `translateTweet`. Guest keys and direct MPP can call it * `GET /x/tweets/{id}/embed` returns the HTML that embeds a tweet on a website, in a light or dark theme * It can hide media and the thread, and leave the widget script out. It costs 1 credit per call * MCP adds it as `getTweetEmbed`. Guest keys and direct MPP can call it * `GET /x/links/resolve` returns where up to 100 t.co links lead, in the order sent * It costs 1 credit per resolved link. A link X does not answer for comes back in `unprocessedUrls`, free * MCP adds it as `resolveLinks`, and guest keys can call it * `GET /x/tweets` takes tweet URLs beside tweet IDs in `ids`, and reads a tweet named twice once * `GET /x/followers/check` takes user IDs beside usernames, and returns `sourceId` & `targetId` * `GET /x/tweets/{id}/subtitles` returns the caption tracks of a tweet's videos, with timed cues & the full text * It costs 1 credit when a track comes back. A tweet without captions costs nothing * MCP adds it as `getTweetSubtitles`, and guest keys can call it * `GET /x/users/{id}/similar` lists the accounts X shows as similar to a profile, in X's order * It costs 1 credit per account. MCP adds it as `getSimilarUsers`, and guest keys can call it * X can list slightly different accounts on each call. Xquik now returns a fuller, steadier list * `GET /x/users/{id}/articles` lists the X Articles a profile published, newest first * Each row is the Article's post with its title, preview & cover. It costs 1 credit per post * MCP adds it as `getUserArticles`, and guest keys can call it * `GET /x/users/{id}/highlights` returns the posts on a profile's Highlights tab * Rows keep X's order. It costs 1 credit per post returned * MCP adds it as `getUserHighlights`, and guest keys can call it * `GET /x/trends/locations` lists the places X offers trends for * Each location has its `woeid`, country, country code & place type * `country`, `placeType`, `q` & `limit` filter the list. It costs 1 credit per location * MCP adds it as `getXTrendLocations` * `GET /x/search/autocomplete` returns the accounts, topics, hashtags & cashtags X suggests for a query * `types` keeps only the kinds you choose. `limit` caps the suggestions * Cashtags carry price, market cap & 24-hour change. It costs 1 credit per suggestion * MCP adds it as `autocompleteSearch` * `GET /x/lists/search` finds public Lists by keyword, at 1 credit per List * Each List has its counts, banner & owner profile. Private Lists stay out * MCP adds it as `searchLists`, and the SDKs as `lists.retrieve_search` * `GET /x/users/{id}/lists` returns the public Lists an account created, in X's order * `GET /x/users/{id}/list-memberships` returns the public Lists that include an account * Each List costs 1 credit. MCP adds them as `getUserLists` & `getUserListMemberships` * `GET /x/spaces/search` finds public Spaces by keyword, host or topic, at 1 credit per Space * It returns live, scheduled & ended Spaces, each once per page. MCP adds it as `searchSpaces` * Space search takes `pageSize`, the posts a page reads: 1 to 100, default 20 * `GET /x/spaces/{id}` returns 1 Space with its state, times, listener counts, hosts & speakers * It takes a Space ID or URL & costs 1 credit per call. MCP adds it as `getSpace` * `GET /x/spaces/{id}/replay` returns an ended Space's recording as an HLS playback URL * A live or scheduled Space answers 409. It costs 1 credit per call. MCP adds it as `getSpaceReplay` * A Space & each Space search row add `sharings`, the posts shared in the Space * Each sharing has the post, the account that shared it & when * A Space cancelled before it started answers 404 `space_canceled` on its replay * `GET /x/broadcasts/{id}` returns 1 X live video with its title, state, times & watch counts * It takes a broadcast ID or URL & costs 1 credit per call. MCP adds it as `getBroadcast` * Guest keys & direct MPP can call it. Reading a broadcast adds no view to it * A broadcast adds `tweetId` & `tweet`, the post that announces it * A Space & each Space search row add `tweet` beside `tweetId` * Job search answers a free 400 `invalid_cursor` when X refuses a cursor * `GET /x/tweets/{id}/retweeters` answers the same free 400 `invalid_cursor` * `GET /x/jobs/search` searches public X job listings by keyword * Filter by location, work type, seniority, employment type & company * Each job returned costs 1 credit. A closed job is left out & free * `GET /x/jobs/{id}` returns 1 job with its description as text & Markdown * `GET /x/jobs/locations` finds places & the location ID job search takes * MCP adds `searchJobs`, `getJob` & `getJobLocations`, and guest keys can call them * `GET /x/places/search` finds X places by name, at 1 credit per place * Each place has the `id` that tweet search takes as `place` * `GET /x/hashflags` lists the custom emoji X shows after hashtags, at 1 credit each * `q` & `activeOnly` filter the hashflags. Each has its image & the time it runs * MCP adds them as `searchXPlaces` & `getXHashflags`, and guest keys can call both **Changed** * `descriptionMarkdown` on `GET /x/jobs/{id}` numbers ordered lists 1, 2, 3 and keeps line breaks * It escapes characters that would read as Markdown or HTML, such as `[`, `*` & `<` * `GET /x/communities/{id}/tweets` opens its first page with the pinned post, flagged `isPinned` * `GET /x/communities/suggested` asks up to 2 more times when X suggests none. An empty list stays free * `GET /x/trends` takes a `country` or a `location` name, such as `location=Istanbul` * The response returns the place it read as `location`. `woeid` still wins * A write's `402` adds `top_up_url` & `dashboard`, which say where to add credits * `404 user_not_found` adds `reason`: `not_found` or `suspended` * A suspended account answers "X suspended this account. Choose another account." * Account responses add `connectedAt`, when the account last connected or reconnected * `X-Xquik-Busy-Reason` is now `admission` or `reads` * `GET /x/users/batch` takes `usernames` in place of `ids`, with or without `@` * Send up to 100 usernames. The ID lists then name the usernames as sent * `usernames` also takes profile links, such as `x.com/nasa`. A link is named by its username * The text `undefined` or `null` in `usernames` now answers 400 & costs nothing * A filtered `Latest` Tweet search adds `readBackTo`, how far back it has read * 7 account lists now page with a cursor instead of stopping at 100 or 200 rows * They are monitors, keyword monitors, webhooks, webhook deliveries, API keys, styles & support tickets * Each takes `limit` & `cursor`, and answers `hasMore` & `nextCursor` * A 1st page keeps its size & order, so existing clients see the same rows * Xquik keeps queued events until all deliveries finish * Events expire 30 days after Xquik creates them * `GET /credits` adds `auto_topup_stopped` & `GET /account` adds `creditInfo.autoTopupStopped` * `true` means 3 declined charges in a row on one card stopped automatic top-up * An event keeps the monitor label from when Xquik created it * Renaming a monitor no longer changes past events or their retries * Each retry of a webhook delivery sends its first attempt's exact body * Tweet & DM `text` now reads as x.com shows it * X's `&`, `<` & `>` arrive as `&`, `<` & `>` * `displayTextRange` & entity `indices` count the decoded text * A repost row now shows its original post's counts, as x.com does * That covers likes, replies, reposts, quotes, bookmarks & views * X's own repost counts were 0, apart from reposts & a smaller view count * Tweet search reads `since:`, `until:`, `sinceDate` & `untilDate` as UTC days * x.com uses your time zone, so its day edges can differ * `untilDate` is inclusive: `untilDate=2026-03-01` returns that day's posts too * Followers come newest first. x.com puts some promoted followers first * A reply read with `sinceTime` & `scope=direct` stops X's newest-first view once it is past your window * `diagnostic.strategiesAttempted[].stopReason` says `window_passed` for that view * Once that view & both window searches end, other views stop with `window_covered` * Such a read ends sooner, with `sourcesEnded: true` * Tweet search drops `include:replies`, which X refuses. Search shows replies anyway * It reads `include:retweets` as `include:nativeretweets` * `searchedQuery` then shows the query X searched * A write retried with the same `Idempotency-Key` returns the first answer again * It keeps the first status, body & `Retry-After`, with `idempotent: true` added * A rejected post's retry now answers `422` like the first answer, not `200` **New** * `DELETE /x/dm/{userId}/messages/{messageId}` deletes a DM from your account's side * The other person keeps it. It costs 10 credits, only when X deletes it * MCP adds it as `deleteDm`, and the SDKs as `dm.delete_message` * Profiles add `accountBasedInUnavailable: true` when X withholds `accountBasedIn` from a read * Those profiles omit `accountBasedIn`. Treat the label as unknown & retry later * `accountBasedIn: null` still means X shows no label **Changed** * Native MCP adds `readSavedResult`, which reads a saved oversized answer in free pages * Oversized answers up to the 2 MiB MCP response limit now save, up from 256 KB * An oversized answer's `retry` names only tools your MCP catalog lists * Tweet search drops an unmatched quote, which X refuses, & returns rows * `searchedQuery` then shows the query X searched * A search X refuses returns `424 search_unavailable`, never an empty page * User Tweets, user replies, quotes & List Tweets list `keywords`, `place`, `placeCountry`, `pointRadius`, `boundingBox` & `advancedQuery` * User Tweets, user replies & quotes also list `listId`, and Tweet search lists `keywords` * A user timeline with a search filter such as `keywords` filters every Tweet, not only older ones * `max_results` sets the page size wherever `limit` & `count` do * `maxItems`, `max_items` & `per_page` set it too * Tweet search `queryType` takes any case & the aliases `type`, `search_type`, `product`, `category` & `section` * `Photos`, `Videos` & `Media` search `Latest` with that media type. `People` & `Lists` answer 400 with the route to use * Direct replies stop listing 8 search filters they never applied * Those are `cardName`, `source`, `excludeSource`, `geocode`, `near`, `within`, `safe` & `news` * List Tweets & direct replies take the legacy `after` cursor alias on every page * Community search applies `withinTime` * Events prefer `monitorId` to `monitor_id`. A monitor filter without a usable ID lists no events * A cursor ignores parameters its route doesn't read, so dropping one keeps it * A post X never published fails as `x_post_not_sent`. You pay nothing & can post again * Writes X never confirms end `expired` & charge nothing. Check your X account before you retry * X's 500, 504 & 408 after the send keep a write `pending_confirmation` as `x_write_ambiguous` * Pending answers say `retryable: false` * Before the send, those give `x_transient_error` & are safe to retry, not `x_write_failed` * X refusals that state a reason return `422 x_rejected`. They used to return `500 x_write_failed` * X's refusal of an image's size returns `422 x_image_invalid` * Avatar & banner uploads take images up to 15 MiB. Xquik fits them to X's limits * Post images over 5 MiB shrink to fit. GIFs & videos go to X unchanged * An unreadable image returns `415 unsupported_media_type` * An image Xquik can't shrink enough returns `413 media_too_large` * OpenAPI lists 413 & 415 on media writes, plus 3 media error codes * Post & user path params take URLs & handles. See [path IDs](/api-reference/overview#path-ids) * Post params take a post ID or an x.com or twitter.com post URL * User params take a user ID, a username with or without `@`, or a profile URL * Writes look up a username first. An unknown username returns `422 x_target_not_found` * Other input returns `400 invalid_tweet_id` or `400 invalid_user_id` * `GET /x/users/{id}` returns `invalid_user_id` where it returned `invalid_username` * Writes X refuses outright return `422 x_rejected` & charge nothing * `POST /x/communities` returns `422 x_account_feature_required` when the account can't create communities * Both used to look like unconfirmed writes * MCP `execute` takes paths with or without `/api/v1`, such as `/x/users/nasa/follow` * `GET /x/users/{id}/tweets` & `GET /x/users/{id}/replies` now list `sinceTime`, `untilTime` & `mode` * Both routes already applied them. `sinceTime` keeps its exact second & `untilTime` drops it * A time without an offset reads as UTC on every route that takes these bounds * An offset whose `+` arrived unencoded, such as `21:20:41 02:00`, reads as `+02:00` * Community Tweets, List Tweets, thread & batch users now list every filter they applied * Quotes list `includeReplies`, and bookmark folders list `cursor` * Likes, media & mentions drop search operators such as `cardName`. They never applied them * Replies to a post that is unavailable on X return `422 x_reply_target_unavailable` * Some used to return `500 x_write_failed` * Xquik sends nothing & charges nothing. Check the post on x.com or reply to another post * `GET /draws`, `GET /drafts` & `GET /events` return `400 invalid_input` for an unknown cursor * Before, they returned the first page again * Send the last `nextCursor` you received, or omit the cursor to start over * An empty cursor still returns the first page * Replies, quotes, retweeters & thread of a post X does not have return `404 tweet_not_found` * They used to return 200 with an empty page. `GET /x/tweets/{id}` already returned this 404 * Only an empty first page checks the post. A post without engagement still returns 200 * A 404 costs no credits * Each account now gets 500 reads a second, up from 300 * Writes still allow 120 a minute & deletes 60 a minute * `429 x_rate_limited` messages now name the wait from `Retry-After`, such as about 15 minutes * Without a known wait, the message says to try again later * Writes X refuses as possibly automated return `422 x_automation_refused` * They used to return `429 x_rate_limited` * Waiting does not change X's answer. Send the next request now with a new `Idempotency-Key` * If X keeps refusing, check the account on x.com. Vary your text or pace * These refusals cost nothing * `429 x_rate_limited` now covers only X's rate limits. Wait for `Retry-After` when present * `GET /x/tweets/{id}/replies` with `sort=latest` or `sort=oldest` sets `orderPartial: true` when some rows come from X views without date order * Rows stay sorted by date & keep the same status. Replies may be missing between them * Nothing marked such answers before. Retry later for a gap-free order * When X's Recent view ends early or comes back signed out, Xquik reads the rest from X search's Latest results * `GET /x/tweets/{id}/replies` applies `sort` on automatic pages too * Those pages used to list the newest replies first, whatever the `sort` * Without `sort`, automatic pages still list the newest first. Standard pages ignore `sort` * `POST /draws` reads every direct reply & retweeter X shows * It used to read X's first page of replies & at most 2,000 retweeters * Entries are direct replies to the post. Replies to other replies no longer enter * A draw on a busy post takes about 1 minute. Keep the request open for 100 seconds * `GET /x/tweets/{id}/retweeters` no longer ends the list early when X refuses a page * Xquik reads that page again * `POST /draws` returns `424` when X stops the reply or repost read early * It used to draw from the entries read so far * Retry after `replies_incomplete` or `retweeters_incomplete` * `draw_too_large` means the post has more entries than 1 draw reads. Contact support * Xquik saves & charges nothing for these draws * Bookmark folder reads on accounts without X Premium return `424 bookmark_folders_unavailable` * This covers `GET /x/bookmarks/folders` & `GET /x/bookmarks?folderId=...` * They used to return an empty list. The message says to upgrade on X or read all bookmarks * These answers cost nothing * Search Tweets returns `424 search_unavailable` at once for `filter:vine`, `filter:consumer_video`, `filter:pro_video`, `filter:news` & `retweets_of:` * X no longer searches these. `news=true` gets the same answer * The message says what to change, such as the term to remove * A query X refuses as a whole also gets `424 search_unavailable`, such as one over 512 characters * These answers cost nothing * Follower, following, verified follower, retweeter & user search pages take `enrichProfiles=true` * Each row then carries the full profile, with fields such as `accountBasedIn` * The `x-xquik-profile-enrichment` header counts the rows that carry it * Extractions of these rows take `enrichProfiles` & add the profile at `enrichmentData.profile` * The price per row stays the same * Complete-mode replies return 200 after reading a whole `sinceTime` or `untilTime` window * Replies outside the window no longer turn that answer into `424 replies_incomplete` * Complete-mode replies add `diagnostic.sourcesEnded` * `true` means every source ended. X hides the other replies * A `424 replies_incomplete` with `sourcesEnded: true` returns the same rows on retry * Its message says X hides the other replies instead of asking for a retry * Complete-mode replies find replies from new & small accounts more reliably * `diagnostic.strategiesAttempted[].stopReason` adds `partial_view` for a view that showed only part of a thread * Such a view keeps `sourcesEnded` false until the read holds every reply the post counts * `withinTime` now filters user Tweets, replies, media, likes & mentions * It also filters Tweet replies & quotes * Posts outside the window no longer come back or cost credits * Use a whole number & `s`, `m`, `h` or `d`, such as `90m` or `7d` * On these routes, other values return `400 invalid_input` * `GET /x/users/batch` returns the profiles it read when some IDs fail * IDs that failed to load come back in `failed_ids`. Retry them * Failed IDs cost nothing * Draws charge follow checks only for authors who pass your other filters * `POST /draws` accepts an optional `Idempotency-Key` header * An exact retry returns the original draw with `Idempotency-Replayed: true` & charges nothing * A retry while the first request runs returns `409 idempotency_in_progress` * Another body with the same key returns `409 idempotency_conflict` * A draw that fails its save or final charge stores nothing & charges nothing * The dashboard sends 1 key per draw, so a resubmit returns the saved draw * Webhook retries have no attempt limit. They continue until your endpoint returns `2xx` * `410 Gone` retries like any other failure * A failing endpoint gets 1 waiting delivery at a time, at least every 15 minutes * When that delivery succeeds, Xquik starts sending `pending` deliveries * Resume Webhook & `isActive: true` start sending waiting deliveries at once * Without them, a rejected delivery can wait up to 7 days after a fix * Only a paused or deleted webhook marks deliveries `exhausted` * Resume within 30 days of an event to receive its held delivery * Return `2xx` for deliveries you skip. A rejected one can delay new ones up to 15 minutes * Private reads use only your connected X account * They cover bookmarks, bookmark folders, notifications, home timeline, likes, likers & mutual followers * Without one, these routes return `424 account_required` * They return `503` while your connected X accounts are busy * `favoriters` & `user_likes` extraction jobs need a connected X account * Without one, the job fails with `errorMessage` set to `errorXAccountRequired` * Guest keys no longer cover likes, likers & mutual followers * Guest keys get `403 forbidden` on these 3 routes * Guest wallets now prepay 30 GET routes * Small conversation searches return more complete results by default * Reply Scraper 1.2.65 reports completion more accurately * Reply reads return more replies within your requested limit * Reply Scraper 1.2.64 recovers replies after a stall & keeps collected rows * Dated author searches with `include:nativeretweets` or `-filter:nativeretweets` return more complete results * Searches requiring protected authors return an access error before collecting results * Tweet Scraper 1.12.99 returns more Latest search results * Multi-author Tweet searches return more complete results * Pagination docs now explain cursor retries & first-page limits * Tweet search keeps exact ID boundaries across pages * `sinceId` excludes its ID. `maxId` includes its ID * MCP reports missing response-field reads without changing returned values * MCP search keeps response descriptions and date-only string types * Profile, list, and quote reads keep explicit reply filters * Image previews resize AVIF sources again * MCP search keeps shared response schemas in `spec.components.schemas`, reducing repeated contract data * Extraction results retain attached public media for mixed tweet, reply, quote, thread, and profile-media sources **New** * Accountless guest wallets prepay 33 GET routes through a scoped `paid_reads` key * `POST /guest-wallets`, `GET /guest-wallets/status`, and `POST /guest-wallets/topups` support hosted checkout after confirmation without an account * Active guest keys expose the same 33-route read-only catalog through API MCP * OAuth discovery advertises claimed `service_auth` agents with revocable access tokens and identity assertions * API MCP v2.6.0 supports MCP `2026-07-28` through `server/discover` * Modern discovery and tool catalogs include private cache hints * `POST /compose` now returns 7 source-specific `radarRecommendations` with endpoints, use cases, and drafting guidance * Reddit Radar items now expose available post text, destination URLs, media, public engagement signals, estimated vote counts, and post state * `GET /x/account-connection-attempts/{id}` reports a tracked connection as pending, successful, failed, or waiting for an email code **Changed** * `GET /x/tweets/search` accepts `pageSize` as a `limit` alias * API MCP v2.6.58 gives `execute` calls 55 seconds of bounded wall time, documents compact result projection, and keeps completed request diagnostics after a timeout * `GET /x/accounts` supports cursor pages through `limit` and `cursor` * The best-practice contract paginates X accounts by default. Legacy responses retain up to 10,000 accounts * Extraction estimates and jobs now default `resultsLimit` to 10,000. Native MCP schemas retain OpenAPI defaults * API MCP v2.6.43 advertises immutable tool catalogs and disables change subscriptions * API MCP v2.6.42 adds OpenAPI-native tools at `https://xquik.com/mcp?codemode=false`. Code Mode remains the default * API MCP v2.6.41 removes the extraction row threshold from OpenAPI routing metadata * API MCP v2.6.40 uses `docs`, `search`, and `execute`. Tool descriptions contain mechanics, not endpoint choices * Modern MCP calls are request-scoped and require no initialization session * Stateless 2025-era MCP clients remain compatible at the same endpoint * Field guidance now documents REST naming exceptions and MCP normalization * Export guidance now separates file formats from downstream column selection * Account connection email-code challenges now return immediately as a resumable `202 requires_email_code` response * `POST /x/accounts` can return a durable `202 pending` response with `Location` and `Retry-After`. Follow that attempt instead of resending credentials * Account connection failures now expose the exact cooldown through `retryAfterMs` and `Retry-After`. Retry controls show a live countdown * Direct MPP now covers 7 fixed-price `charge` operations. Result-sized reads use guest or full account credits * Follow checks and article reads cost USD 0.00075 per direct MPP call * Every response after an accepted MPP payment includes `Payment-Receipt`, including non-2xx responses * The 26 non-MPP anonymous paid reads return `401` with `WWW-Authenticate: Bearer` and a guest wallet action * The 7 direct MPP reads return `402` with `WWW-Authenticate: Payment` and the same guest action * Reddit Radar now lists only items with full details * Startup growth Radar items now add founder, company, acquisition, margin, and revenue-efficiency fields when available * A successful Compose score now points to `GET /x/accounts` and `POST /x/tweets`, while retaining the one-click `intentUrl` **Removed** * Obsolete public account location fields **Changed** * From July 21, 2026, X write endpoints require an `Idempotency-Key` header * Requests without it return `400 missing_idempotency_key`. Xquik does not charge them * A key that is not 1 to 255 visible ASCII characters returns `400 invalid_idempotency_key` * Send 1 unique key per intended write. Reuse it only to retry that same write **New** * `POST /x/tweets` - added `media` field (1-4 image URLs). `text` is now optional when media is present * `POST /x/accounts/{id}/reauth` - added `email` parameter * `POST /x/media` - now accepts `image/webp` and `image/avif` (allowed: AVIF, GIF, JPEG, PNG, WebP, MP4) * `POST /x/accounts` and `POST /x/accounts/{id}/reauth` responses now include a `health` field (enum: `healthy`, `locked`, `needsReauth`, `recovering`, `suspended`, `temporaryIssue`). The same field now appears on `GET /x/accounts` list and detail responses * New error code `login_cooldown` (429) for `POST /x/accounts` and `POST /x/accounts/{id}/reauth`, with `reason` and `retryAfterMs` in the body and a `Retry-After` header * All 17 write endpoints (under `/x/tweets`, `/x/users`, `/x/dm`, `/x/profile`, `/x/media`, `/x/communities`) now document standardized error responses: `403 account_needs_reauth`, `403 account_restricted`, `422 write_rejected`, `429 rate_limited_by_x`, `503 transient_error` * `GET /radar` items now expose 15 metadata fields per source (was 8): `mrr`, `growthPercent`, `last30Days`, `total` (required), `customers`, `activeSubscriptions`, `onSale`, plus optionals `askingPrice`, `country`, `growthMrrPercent`, `multiple`, `paymentProvider`, `rank`, `category`, `xHandle` * Expanded Machine Payments Protocol (MPP) pay-per-use to 31 endpoints (tweets, communities, lists, user data) * New framework guides: LangChain, Pydantic AI, CrewAI, Mastra, Google ADK, Microsoft Agent Framework **Changed** * Billing migrated to credits-only model (usage-quota removed) * `GET /account` response: `creditInfo` replaces `currentPeriod` * `POST /extractions/estimate` response: `creditsRequired` / `creditsAvailable` replace `usagePercent` / `projectedPercent` * Monitor delivery changed. The public API contract did not change. * `POST /x/tweets/{id}/like` and `POST /x/tweets/{id}/retweet` are now end-to-end idempotent. Calling either on an already-acted-on tweet returns 200 success. The earlier silent rejection produced a 500 plus an account cooldown. **Removed** * Xquik no longer emits event types `follower.gained` and `follower.lost`. Valid event types: `tweet.new`, `tweet.quote`, `tweet.reply`, `tweet.retweet` (plus `webhook.test`). * Error code `shadow_account` (422) removed from `POST /monitors` * Error code `stream_registration_failed` (502) removed from `POST /monitors` (monitor activation is now asynchronous) * Telegram bot endpoints and integrations pipeline endpoints removed **New** * MPP expanded from 16 to 31 eligible endpoints * Webhook delivery signing now uses HMAC-SHA256 with `X-Xquik-Signature` header **Changed** * OAuth 2.1 discovery metadata now served from `/.well-known/oauth-authorization-server` * Rate limit responses include `Retry-After` header This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.