# Threads Search API (https://www.socialcrawl.dev/platforms/threads/search)

> Searches Threads for posts matching a keyword query. Returns a list of matching posts with text content, like counts, author info, and creation timestamps. Posts filed under a Threads topic tag also carry that tag name at `post.ext.topic_tag` (e.g. `bali`, `dokter gigi`) with its stable id at `post.ext.topic_tag_id`, so you can group or filter results by topic client-side; both are absent on untagged posts, which are the majority. Threads exposes no tags-only search mode upstream, so this stays a keyword query. Search rows carry the same media shape as the other post lanes: carousel posts return every slide as an array in `content.media_urls`, video posts return a playable video URL there with the cover frame in `content.thumbnail_url`, and a quote post's attached content is recoverable at `post.ext.quoted_post`. `post.ext.quote_count` and `post.ext.reshare_count` complete the reshare picture that `engagement.shares` only starts. `engagement.views` and `post.flags.pinned` are null on a plain call, because Threads publishes them only on `/v1/threads/post`; send `include=engagement` and the first 20 posts are looked up there by permalink in the same call, and only those two leaves (plus a missing display name) are copied onto each row. You are billed 1 credit for each post whose view count or display name was filled. Threads publishes a view count only once it has one, so a post that is minutes old usually comes back without one: that row still gets its pinned flag, and it is free. The join adds 3 to 12 seconds (never more than 12), and a `data.hydration` block reports the rows looked up, filled and served from cache and the credits held and kept; a post the lookup could not read at all is refunded and named by `data._warnings: ["engagement_partial"]`. It is refused beside a `limit` above 20: page with the cursor instead, 20 posts a call. Each response is one result window (typically 15-20 posts); page deeper by sending back pagination.next_cursor as cursor, or set limit (1-100) to have the API walk multiple windows server-side until it has collected at least that many unique posts, billed per window consumed (1 credit each, unused window budget refunded). limit is a collection target rather than a page size, so a response usually carries somewhat more than you asked for; nothing you were billed for is trimmed away. Those windows run sequentially at roughly 3.4 seconds each, so limit buys posts with wall clock as well as credits: measured 07/09/2026 cache-cold, a single window is 3.5s while limit=50 is 10.2s and limit=100 is 19.5s. Set your client timeout above the limit you ask for, or page with the cursor instead. start_date and end_date bound the search to a period; the cursor walk stays inside those bounds. Multi-word queries are handled for you: Threads matches a search against the topic tags it already knows rather than scoring individual terms, so a longer phrase it does not recognise returns nothing at all rather than fewer results. When that happens the query is automatically relaxed into its adjacent word pairs (for example `cari vendor mesin kopi jakarta` becomes `cari vendor`, `vendor mesin`, `mesin kopi` and `kopi jakarta`), those searches run in parallel, and the results come back merged, de-duplicated and ordered by how many of your terms each post contains. If every pair still finds nothing, or the query is an unspaced Chinese, Japanese or Thai phrase, its single words are searched next and only posts that contain your whole phrase, both words of a two-word query, or at least two words of a longer one are kept. You are billed 1 credit for each relaxed search that returned posts and nothing for the ones that did not, so a query that already worked costs exactly what it always did. An expanded response is a single page and carries no pagination cursor, and `data._warnings` names every sub-query that ran. Relaxation respects start_date and end_date: posts outside your window are dropped from every sub-query, and a sub-query left with nothing inside the window is not billed. It also never fires on a request that carried a cursor, so the last hop of a pagination loop is the honest end of the walk rather than a surprise page of loosely related posts. Send expand=false to search the exact phrase only.

TL;DR: `GET /v1/threads/search` costs 1 credit per result window. A call with no `limit` is one window and costs exactly 1. With `limit` set the walker consumes `ceil(limit / 15)` windows - 4 for limit=50, 7 for the 100 maximum - holds that many up front and refunds every window it did not need. Query relaxation can add up to 4 more windows on a page-1 request, billed only for the relaxed searches that returned posts and refunded for the ones that did not; send `expand=false` to hold none of them. When your balance, or the API key's limit, covers the call without them but not with them, it runs as `expand=false` instead of answering 402, and `data._warnings` says so. A request carrying a cursor never relaxes, so a pagination loop's last hop costs nothing extra. With `include=engagement` the call also holds 20 credits and keeps 1 for each of the first 20 posts whose view count or display name it filled from a fresh lookup; posts served from the post lookup's cache (10 minutes) are free, a post Threads has published no view count for is free (it still gets its pinned flag), and the rest is refunded Relevance against your query (by default, or with relevance=) is free. Adding relevant_to= holds 4 extra credits and refunds down to 1 credit per started 25 rows that were newly judged: rows already judged for the same topic are free, a page where nothing could be judged refunds the whole extra, and a cached page is free. The default labels (sponsored, intent, niche) are free. label=mention, quality, ai_style, ai_written, injection, or label=intent with offer=, holds 4 extra credits and refunds down to 1 credit per started 25 posts that were newly judged: posts already labelled are free, a page where nothing could be judged refunds the whole extra, and a cached page is free. With `max_pages`, `max_items` or `max_credits`, each page walked is billed at this price (a cached page is free), and `max_credits` caps the total. With `seen`, a page's price falls with its share of rows you already received under that id; credits for joined data (`include=`) are never discounted. and returns SocialCrawl's unified JSON schema. Single x-api-key auth, 100 free credits on signup.

## Parameters

| Parameter | Required | Description |
| --- | --- | --- |
| `query` | yes | Search keyword or phrase to find Threads posts |
| `start_date` | no | Inclusive start of the search period (YYYY-MM-DD). Pagination never walks past it, and a relaxed query drops any post older than it. Threads treats the date pair as a ranking hint rather than a filter, so the window is enforced on our side: a post we cannot date is treated as outside it. |
| `end_date` | no | Inclusive end of the search period (YYYY-MM-DD). Pagination starts here and walks back in time, and a relaxed query drops any post newer than it. |
| `trim` | no | Ask the upstream for a stripped record. This is a real shape change, not just less whitespace: a trimmed row carries only the id, text, shortcode, like count, timestamp and author, so media, reply count, share count, view count, topic tag, pinned flag and quoted post all come back null. Same 1 credit either way, so send it only when the text and the like count are all you need. |
| `cursor` | no | Opaque cursor from a prior response's pagination.next_cursor, passed back verbatim, to fetch the next (older) result window. It carries your date bounds, so paging cannot escape the period you asked for. A request that sends a cursor is never query-relaxed. |
| `limit` | no | Collect AT LEAST this many unique posts in one call (1-100). This is a collection target, not a page size: the API walks whole result windows server-side until it has collected this many (or the query runs dry), so the response usually carries somewhat more than you asked for and never fewer unless the query ran out. Nothing you paid for is trimmed away. Billed 1 credit per window consumed (about 15-20 posts each), with the unused window budget refunded. Omit for a single window. BUDGET FOR THE WALL CLOCK: the windows run one after another, at roughly 3.4 seconds each, so a high limit is a long request. Measured on production 07/09/2026, cache-cold: no limit 3.5s for 20 posts, limit=30 7.3s for 37, limit=50 10.2s for 55, limit=100 19.5s for 111. If your HTTP client defaults to a 10-second timeout, limit=50 and above will not fit inside it. Either raise the client timeout, or ask for a smaller limit and page with pagination.next_cursor, which costs the same credits for the same posts. |
| `expand` | no | Set to false to search the exact phrase only. Defaults to true, which relaxes a query Threads does not recognise into its adjacent word pairs and merges the results, billing 1 credit per relaxed search that returned posts inside your date window. Ignored on a request that carries a cursor. When the credits you have cover the call with expand=false but not as sent, the call runs with expand=false instead of answering 402, and data._warnings says so. |
| `include` | no | Send `engagement` to fill `engagement.views` and `post.flags.pinned` on the first 20 posts by looking each one up on `/v1/threads/post` in the same call. 1 credit per post whose view count or display name was filled from a fresh lookup (20 at most); posts served from that lookup's cache (10 minutes) are free, and so is a post Threads has published no view count for yet, which still gets its pinned flag. Adds 3 to 12 seconds, never more than 12. One token only: `engagement`. Not accepted with a `limit` above 20; page with the cursor for more. (engagement) |
| `relevance` | no | Optional. Without this param every row already carries computed.relevance against your query, free ({ p, sense, depth, spam }: is this row about what your query means, or a different thing that shares its words?), and nothing is dropped or reordered. score asks for it explicitly and waits for every row; filter also drops the rows that are not about your query, and lists their ids in data.relevance.dropped_ids. Your query is the topic; nothing to configure. A row that could not be judged is never dropped and carries relevance: null. Pagination is unchanged, so a filtered page can hold fewer rows. Free with your query as the topic; with relevant_to it adds 1 credit per started 25 newly judged rows (rows already judged for the same topic are free, and so is a cached page). (score \| filter) |
| `relevance_threshold` | no | Optional, only with relevance. The probability (0 to 1) a row must reach to be kept by relevance=filter. Default 0.5. Lower keeps more rows, higher keeps fewer. |
| `relevant_to` | no | Optional, only with relevance. Up to 200 characters describing what you mean, used as the topic instead of the query. Use it when the query is ambiguous, for example query=cleopatra with relevant_to=Cleopatra, the IGT slot game. |
| `judgments` | no | Optional, on (the default) or off. By default every row gains free SocialCrawl judgments (computed.labels, and computed.relevance on search endpoints), reported in data.labels (mode default) and data.relevance (origin default), each with a status (complete, partial or skipped) and pending: the rows still being judged when the page was sent, which carry null now and are filled on your next call or cached read. Default judgments never add credits, never change an existing field, and never drop or reorder a row. off returns the page exactly as before, with none of those keys. label=none does the same. (on \| off) |
| `dry_run` | no | Optional. When 1, return a cost preview for this labelled or relevance-filtered request without fetching the page or judging any row. data.estimate reports rows_expected, rows_cached, label_credits_min, label_credits_max and base_credits. 0 credits charged. (1) |
| `fit` | no | Optional. When goal, keep the rows and fields needed for the goal you pass in goal= (plus any that are uncertain, and the first and last), and replace the rest with a stub. data.held_back lists the held ids and a recall id that re-reads the full page from cache at no extra charge. Without this param the page is unchanged. (goal) |
| `goal` | no | Required by fit=goal. What you are trying to do, in your own words, up to 300 characters. |
| `fit_tokens` | no | Optional, only with fit=goal. Soft cap on how much of the page to keep, in tokens. Uncertain blocks and the first and last block are kept even if they exceed it. |
| `label` | no | Optional CSV of SocialCrawl labels to add to every post. Without this param every page already carries sponsored, intent and niche, free; label= adds the labels you name to them (the defaults keep running). sponsored, intent and niche are free when asked for too; mention, quality, ai_style, intent with offer=, and injection add 1 credit per started 25 newly judged posts. judgments=off (or label=none) turns the default labels off. sponsored: is the post a paid or gifted promotion (p, 0 to 1), did it carry a disclosure marker such as #ad or 광고 (disclosed), is it likely paid with no marker (undisclosed), and which of the accounts it mentions does it promote (brand, or null). These are signals to review, never a finding. mention (needs brand=): is the post about that brand rather than something that shares its name (about_brand, 0 to 1), how it feels about the brand on five levels (sentiment_level 0 to 4 and sentiment_score 0 to 1, null when the post is not about the brand), is it sarcastic, which aspect it talks about (taste_or_quality, price_or_value, availability_or_delivery, health_or_safety, advertising_or_campaign, customer_service, none), and did the author buy or use it (first_hand). intent: what the author is mainly doing (label: asking_for_recommendation, comparing_options, switching_away, complaining, promoting, news_or_discussion, other, or null when unsure, with confidence), whether they read as a potential buyer rather than a seller (buyer, seller), how pressing the need is (urgency, 0 to 3), and, when you pass offer=, whether your offer would plausibly help them (fits_offer). niche: which of the 33 niches of the published taxonomy sc-niche-v1 the caption belongs to, or personal_no_niche, or other (label, confidence, taxonomy), with label null when the caption is too thin to tell or the pick is unsure. It reads the caption only, not the video. quality: how much checkable detail the caption carries (fact_density 0 to 3), whether it mainly asks for likes, replies, shares, follows or tags (engagement_bait, 0 to 1), whether it is written to provoke anger as a way to get engagement (rage_bait, 0 to 1; about the writing, never the side taken), whether it only repeats someone else's news or view (secondhand, 0 to 1), and what the post is mainly doing (post_aim: inform, opinion, sell, entertain, provoke, other). exclude=engagement_bait drops posts whose engagement_bait is 0.8 or higher. ai_style: writing patterns common in AI-generated text, judged on the text alone. It does not say who wrote the post, and a person who writes in this style is flagged too. Six patterns are judged, each a probability (0 to 1): binary_contrast ("This is not X. It's Y."), faux_insight ("Here's what nobody tells you" and throat-clearing openers), colon_reveal ("The best part: it learns."), profound_kicker (ends on an aphorism or mic-drop line), puffery ("marks a pivotal moment", "highlighting the importance of") and rhetorical_setup ("What if I told you", a question the post answers itself, "Not a X. Not a Y. A Z."). Four are counted in the text: slop_words (words and phrases such as delve, leverage, robust, game changer, at the end of the day), em_dashes, emoji_bullets (lines that start with an emoji or a symbol bullet) and one_line_paragraphs (paragraphs that are one short line). pattern_count is how many patterns are present (a judged pattern at 0.8 or higher, slop_words 1 or more, em_dashes 2 or more, emoji_bullets 2 or more, one_line_paragraphs 5 or more) and markers lists their names, comma-separated. A post with less than about eight words of text is only counted, with every judged pattern 0. The judged patterns read the first 400 and the last 150 characters of a long post, and the counts read the whole post. exclude=ai_style drops posts with 2 or more patterns. ai_written: how likely the post was written by an AI model, from SocialCrawl's own text classifier, reading the opening of the post (about 1,000 characters): p (0 to 1, calibrated), level (likely (p 0.95 or higher), possible (p 0.71 or higher) or unlikely) and model (the model name; a new name means a new scale). A post under 8 words is too short to read: p null, level too_short. It is a probability, not proof: a person can be flagged and an AI-written post can pass, and the shorter the post the less certain it is. exclude=ai_written drops likely posts. injection flags text that addresses an AI system and tries to direct it (flagged, p); it never drops or rewrites a row. A post that could not be judged carries labels: null. Posts already labelled for anyone are free, and so is a cached page. data.labels reports what was judged and billed. |
| `exclude` | no | Optional CSV (engagement_bait, ai_style, ai_written). engagement_bait, only with label=quality, drops posts whose engagement_bait probability is 0.8 or higher. ai_style, only with label=ai_style, drops posts that show 2 or more of its writing patterns (pattern_count 2 or higher). ai_written, only with label=ai_written, drops posts whose level is likely (p 0.95 or higher), or whose p reaches your ai_written_threshold; a post too short to read is never dropped. Dropped posts are listed in data.labels.dropped_ids. A post that could not be judged is never dropped. |
| `ai_written_threshold` | no | Optional, only with label=ai_written and exclude=ai_written. The p (0 to 1) at which exclude=ai_written drops a post. Default 0.95 (level likely). 0.71 also drops possible posts: more AI posts go, and more posts by people with them. |
| `brand` | no | Required by label=mention, ignored otherwise. The brand or product the posts are judged against, up to 300 characters, for example brand=Buldak. Without it label=mention is skipped with the warning label_mention_needs_brand and is not billed. |
| `brand_description` | no | Optional, with label=mention. One plain sentence saying what the brand is, for example brand_description=Samyang's spicy instant noodle brand. Use it when the name is also an ordinary word or another company's name. |
| `offer` | no | Optional, with label=intent. One or two plain sentences saying what you sell, up to 300 characters, for example offer=A web design agency that builds websites for small businesses. Adds fits_offer to every post; without it fits_offer is null. |
| `label_evidence` | no | Optional, only with label=. When 1, every labelled row also carries computed.labels_evidence.<preset> = { quote, sentence_index }: the sentence in the row that most clearly shows the label, copied verbatim. Absent or null when no single sentence shows it. (1) |
| `since` | no | Only rows published on or after this date: YYYY-MM-DD (midnight UTC) or an ISO 8601 timestamp. Rows older than it are left off each page. This list is not in date order, so the walk does not stop at the date: keep following `next_cursor`, and each page costs what a page costs even when every row on it is filtered out. A row with no date is kept. `published_after`, `start_date`, `from_date`, `publish_time` (`yesterday`, `this-week`, `this-month`, `last-3-months`, `last-6-months`) and `max_age_days` are accepted and applied as `since`. |
| `until` | no | Only rows published on or before this date: YYYY-MM-DD (that whole UTC day is included) or an ISO 8601 timestamp. Newer rows are left off each page and the walk does not stop, so older pages follow. A row with no date is kept. `published_before`, `end_date` and `to_date` are accepted and applied as `until`. |
| `max_pages` | no | 1 to 50 (default 1, or 50 when `max_items` or `max_credits` is sent). Walk up to this many pages in one call and return the rows from all of them. Each page walked is billed exactly as one call to this endpoint (a cached page is free) and counts against your rate limit; its charge and any refund are recorded under the page's own request id, listed in `data.walk.request_ids`. The walk stops early at the last page, or when the request's time budget runs low; `data.walk.stopped` says why (`end`, `max_pages`, `max_items`, `max_credits`, `time_budget`, `page_error`) and `pagination.next_cursor` continues from where it stopped. |
| `max_items` | no | 1 to 5000. Get more in one call: keep walking pages until the response holds at least this many rows, then stop with `data.walk.stopped: "max_items"`. The last page is returned whole (it was billed), so `data.items` can hold up to one page more than this, and `pagination.next_cursor` resumes right after it without skipping a row. Walks up to 50 pages unless `max_pages` says fewer; each page is billed as one call (a cached page is free). Add `max_credits` to cap the spend. A long walk can stop on the time budget first (`"time_budget"`): send the returned cursor with the same params to continue. |
| `max_credits` | no | 1 to 100000. The most this call may spend, across every page it walks. A page starts only while the credits already spent plus that page's price (or, when higher, the most any page of this walk has cost) still fit, so the call never goes past it; the walk then stops with `data.walk.stopped: "max_credits"` and the cursor to resume. Sent alone, it walks as far as the budget goes (up to 50 pages). A value below the price of one page is a free 400 whose `error.details.page_credits` names that price. Credits a row join (`include=`) adds are known only once a page has run, so page 1 is checked against its own price. |
| `seen` | no | An id you choose (1 to 64 letters, digits, `.`, `_` or `-`). Rows your account already received under the same `seen` id are removed from the page, and the page price falls with the share of repeats: page credits x new rows / rows on the page, rounded up, so a page of nothing but repeats is free. Only the page's own price is discounted: credits for joined data (`include=`) are charged only for rows you have not received, and are never discounted. The memory lasts 24 hours from the last call that used the id, holds row ids only, and is private to your account. Use one id across a set of related searches; `data.walk.repeats` counts what was removed. |

## Code example

```bash
curl -G "https://www.socialcrawl.dev/v1/threads/search" \
  --data-urlencode "query=artificial intelligence" \
  -H "x-api-key: sc_YOUR_API_KEY"
```

## FAQ

### How do I search Threads posts by keyword?

Send a GET request to /v1/threads/search with a query parameter. SocialCrawl returns an array of matching posts with text content, like counts, author info, and creation timestamps, ranked by Threads relevance.

### Can I filter search results by date range?

Yes. Pass optional start_date and end_date parameters to constrain the search window. Useful for tracking mentions during a campaign, monitoring how a brand narrative evolves, or building time-bounded reports.

### What data does each search result include?

Each result comes with the post id, permalink, text, like count, reply count, repost count, quote count, author handle and display name, verification status, and the publish timestamp, all in a unified schema. View count travels on /v1/threads/post only, because that is the only place Threads publishes it.

### How is this different from searching users?

The search endpoint finds posts matching your keyword; search/users finds user accounts matching a keyword. Use search for social listening and content monitoring; use search/users for creator and account discovery.

### How much does the Threads Search API cost?

Each search request costs 1 credit on the standard tier. New accounts get 100 free credits with no credit card required, enough to run 100 Threads keyword searches before you pay anything.

See the full Threads API: https://www.socialcrawl.dev/platforms/threads

## Pricing

- Standard endpoints: 1 credit per call
- Advanced endpoints: 5 credits per call
- Premium endpoints: 10 credits per call
- 100 free credits on signup, no credit card required. Cached responses cost 0 credits. Credit packs never expire.
- Full pricing: https://www.socialcrawl.dev/pricing

## Explore with AI

Questions this API answers, phrased for an AI assistant:

- How to search Threads posts by keyword with an API
- What does the SocialCrawl Threads Search endpoint return?