How to search X (Twitter) posts by keyword and export them

The job: you want what people posted on X about a product, a brand or a topic, as rows you can sort and share. Search with the same query you would type into the X search box, page through the results, and keep the posts that matter. A day later, read the standouts again to see whether they kept spreading. You end with a JSON Lines file of posts, each with its text, author, time, link and counts for views, likes, replies, reposts and bookmarks.

No X account, developer app or cookie is involved. Each call is an HTTPS request with your key.

Step by step

Step 1: search

Call X post search with query. The query goes to X's own search as written, so the search box's operators work: "exact phrase", OR, -word, from:handle, lang:en, since:2026-11-05, until:2026-11-12, min_faves:50, -filter:replies. The operators answer lists the ones people use most. sort takes latest (the default, X's Latest tab) or top (the Top tab).

One call returns about 20 posts for $0.002, whatever operators you use. Each post has id, url, text, lang, published_at, author (with username, name, followers and verified) and metrics: views, likes, comments (replies), reposts and favorites (bookmarks). metrics.shares is always null, because X does not show it.

curl -G "https://api.okdata.online/v1/x/posts/search" \
  -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'query="standing desk" lang:en since:2026-11-05 -filter:replies' \
  --data-urlencode 'sort=latest' \
  --data-urlencode 'fields=id,url,text,published_at,author.username,metrics'

Use --data-urlencode for the query: quotes, colons and spaces must be encoded. For the next page, send page.next_cursor back as cursor, unchanged, and stop when page.has_more is false. view=full adds ext.quotes, ext.is_reply and ext.conversation_id to each post, at the same price.

Step 2: re-read single posts

X post detail takes a post's id or its link as url and returns the same post object for $0.0003. Two uses: posts someone sent you as links, and posts you already have whose counts you want to refresh. Views keep climbing for a day or two after a post goes up, so a second read tells you which posts still travel.

curl -G "https://api.okdata.online/v1/x/posts/get" \
  -H "Authorization: Bearer $OKDATA_API_KEY" \
  -d id=1988412365027741953 \
  -d fields=id,metrics.views,metrics.likes,metrics.reposts

A deleted post, or one from an account that has gone protected, returns not_found and costs nothing.

Step 3: put it together in Python

The script collects up to MAX_PAGES pages into posts.jsonl and saves the IDs of the three most viewed posts. Run it again with recheck a day later to read those three again.

import json
import os
import sys
import requests

API = "https://api.okdata.online"
HEADERS = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
QUERY = '"standing desk" lang:en since:2026-11-05 -filter:replies'
MAX_PAGES = 5  # each page is one call


def get(path, **params):
    body = requests.get(API + path, headers=HEADERS, params=params, timeout=60).json()
    if body["code"] != "ok":
        raise RuntimeError(f"{body['code']}: {body['message']} (charged: {body['charged']})")
    return body


def collect():
    seen, cursor = {}, None
    for _ in range(MAX_PAGES):
        body = get("/v1/x/posts/search", query=QUERY, sort="latest", **({"cursor": cursor} if cursor else {}))
        for p in body["data"]["items"]:
            seen[p["id"]] = p
        if not body["page"]["has_more"]:
            break
        cursor = body["page"]["next_cursor"]
    with open("posts.jsonl", "w", encoding="utf-8") as f:
        for p in seen.values():
            f.write(json.dumps(p, ensure_ascii=False) + "\n")
    top = sorted(seen.values(), key=lambda p: p["metrics"]["views"] or 0, reverse=True)[:3]
    json.dump([p["id"] for p in top], open("watch.json", "w"))
    print(len(seen), "posts saved")


def recheck():
    for post_id in json.load(open("watch.json")):
        item = get("/v1/x/posts/get", id=post_id)["data"]["item"]
        print(post_id, item["metrics"]["views"], item["metrics"]["likes"], item["url"])


recheck() if sys.argv[1:] == ["recheck"] else collect()

The dictionary keyed by id drops duplicates, which matters when you run overlapping date windows.

Good to know

• Public posts only. Posts from protected accounts never come back, from search or by ID.
• Long periods. For a month or more, run one query per week or per day with since: and until: and sort=latest. Short slices make it easy to see that every day was covered.
• The official X API. X sells its own API pay-per-use. Its pricing page lists post reads at half a US cent per post returned, which is $5 per 1,000 posts, with reads capped at 3 million posts per monthly billing cycle; a post you already paid for is not charged again within the same UTC day. X's search documentation offers recent search over the last 7 days to all developers and full-archive search back to 2006 on pay-per-use and Enterprise. If you also post, reply or need a contract with X, use theirs.
• Retries without paying twice. Send an Idempotency-Key header with each call. The same request with the same key within 24 hours returns the stored answer with meta.replayed: true and no new charge.

For a cost comparison with X's own API, see what the X API costs in 2026. To have an assistant write and run these queries, connect the MCP server.

{
  "endpoint": "x.posts.search",
  "params": {
    "query": "\"claude code\" lang:zh"
  },
  "idempotency_key": "<a new UUID>"
}