# Monitor Threads keywords with the Throk API

Keyword monitoring usually takes two steps: use `/post/search` to see **which posts** are talking about it, then `/post/keyword-trend` to see **how the volume changes**.

## 1. Find recent discussions

Search Traditional Chinese posts from the last 7 days that mention 珍珠奶茶 (bubble milk tea), sorted by likes:

```bash
curl -G 'https://premium-api.throk.ai/post/search' \
  -H 'Authorization: Bearer sk-throk-...' \
  --data-urlencode 'keywords=珍珠奶茶' \
  --data-urlencode 'language=zh-Hant' \
  --data-urlencode 'max_age_hours=168' \
  --data-urlencode 'sort=like:desc' \
  --data-urlencode 'size=20'
```

- For the newest discussions, use `sort=timestamp:desc` instead.
- To also get the author, media and repost count, add `include_details=true` (2 credits per post).
- For top-level posts only, without replies: `post_type=thread`.

## 2. See the volume trend

Compare how many times two keywords were mentioned each day over the last 30 days, with daily engagement. 珍奶 is the common short form of 珍珠奶茶:

```bash
curl -G 'https://premium-api.throk.ai/post/keyword-trend' \
  -H 'Authorization: Bearer sk-throk-...' \
  --data-urlencode 'keywords=珍珠奶茶' \
  --data-urlencode 'keywords=珍奶' \
  --data-urlencode 'language=zh-Hant' \
  --data-urlencode 'days=30' \
  --data-urlencode 'include_engagement=true'
```

Response (excerpt):

```json
{
  "results": [
    {
      "keyword": "珍珠奶茶",
      "trend": [
        {
          "date": "2026-09-22",
          "count": 412,
          "like_count": 18230,
          "reply_count": 1640,
          "repost_count": 95,
          "reshare_count": 3120
        }
      ]
    }
  ],
  "credits_consumed": 240
}
```

- `count`: the number of posts (including replies) published that day (Taiwan time) that mention the keyword
- `like_count`, `reply_count`, `repost_count`, `reshare_count`: the total likes, replies, reposts (including quotes) and shares of those posts. These are **cumulative numbers** for each post as of Throk's last update, not the engagement added that day

Days are split at 00:00 Taiwan time. `days` is at most 90.

Use the same `language` for search and trend so the numbers line up. The default is `zh`, which covers all Chinese posts (Traditional, Simplified and Cantonese).

:::caution[The last two days will still go up]
- The last day is today, so it only counts posts so far.
- New posts take a while to be fully collected, so even yesterday, which has already ended, can still go up.
- Engagement is cumulative. Likes and replies on posts from the last few days are still growing.

When comparing trends or writing a weekly report, leave out the last two days, or query again two days later.
:::

## Tips for choosing keywords

Keywords match as whole phrases, so your choice of words directly affects the results. The full rules are in [Search rules](/en/api/guides/search-rules/). These are the points that matter most for monitoring:

- **Send every spelling.** For example `珍珠奶茶` and `珍奶`, or `減肥` and `减肥` (Traditional and Simplified). `/post/search` takes up to 30 keywords at a time, combined with OR.
- **English matches by word.** `AI` won't find "OpenAI". If you want it, add `OpenAI` to the list. English posts (`en`) merge singular and plural, so `app` also finds "apps".
- **Account for Korean particles.** `맛집` won't find "맛집에서". List the common forms too.
- **Avoid single characters.** `茶` (tea) finds every post that contains 茶, so the results get noisy.

## Total volume across spellings

Each keyword's trend line is **counted on its own**, and the same post can appear in several lines, so you can't simply add them up:

- If one post mentions both 珍珠奶茶 and 珍奶, it counts once in each line.
- When one keyword contains another, the shorter one already covers the longer one: the count for `奶茶` (milk tea) already includes every post that mentions 珍珠奶茶.

For example, measured on 2026-09-22: `珍珠奶茶` had 219 posts and `珍奶` had 470, which adds up to 689. The number of posts that mention either one was actually 679. `奶茶` had 2,365 posts, which already include those 219 posts about 珍珠奶茶.

So adding several lines together gives you an **upper bound**: the real total sits between "the largest single line" and "the sum of all lines". When spellings rarely appear in the same post, the sum is very close to the real number.

When you need an exact total, use `/post/search` instead: send all the spellings at once (they're combined with OR, so each post counts once), set the time to the same day and use `size=1`. `metadata.total_count` in the response is that day's post count, for 1 credit per call:

```bash
# Number of posts mentioning 珍珠奶茶 or 珍奶 during all of 2026-09-22 (Taiwan time)
curl -G 'https://premium-api.throk.ai/post/search' \
  -H 'Authorization: Bearer sk-throk-...' \
  --data-urlencode 'keywords=珍珠奶茶' \
  --data-urlencode 'keywords=珍奶' \
  --data-urlencode 'language=zh' \
  --data-urlencode 'from_timestamp=1790006400' \
  --data-urlencode 'to_timestamp=1790092799' \
  --data-urlencode 'size=1'
```

- `from_timestamp` and `to_timestamp` are both inclusive, so one day runs from 00:00:00 to 23:59:59.
- `total_count` shows at most 10,000. If a single day has more, it shows 10,000.

## Get notified about new posts

Don't want to call the API on a schedule? Set up a [webhook](/en/api/webhooks/). When a post with your keyword reaches a likes, replies or views threshold, Throk pushes it to your server.

## Monitor from an AI assistant

With MCP, you can ask an AI assistant such as Claude or ChatGPT to do this for you, for example:

> Use Throk to get the 30-day volume trend for "珍珠奶茶" and "珍奶" in Taiwan, and list the 5 posts with the most likes.

For setup, see [Use in AI assistants (MCP)](/en/api/mcp/).