## tiktok/search

`GET /v1/tiktok/search`

Search TikTok videos by keyword, newest-relevant first, with per-video engagement counts. Paged by cursor.

*Video search*

| Parameter | Required | Description |
| --- | --- | --- |
| `query` | yes | Search term. — e.g. `coffee brewing` |
| `cursor` | no | Opaque cursor from the previous page. |
| `region` | no | Two-letter country code for the source request location. It can affect availability and ranking; it does not filter creators by residence or audience country. Keep it unchanged when continuing. — e.g. `US` |
| `sort_by` | no | relevance (default) or most-liked. |
| `date_posted` | no | Restrict by post date, e.g. past-week. |
| `dry_run` | no | Read a zero-credit estimate without fetching sources, running AI, or reserving credits. Cache status is a snapshot, not a guarantee at execution. — e.g. `1` |
| `judgments` | no | Add bounded AI labels to eligible returned records (default on). off preserves the original data shape. Unclassified records remain present; labels depend on evidence, time, capacity and the existing request budget. No label surcharge. — e.g. `on` |
| `label` | no | Set none to disable labels; default enables the endpoint’s label set. Equivalent to judgments=off. — e.g. `none` |
| `label_evidence` | no | Include exact supporting sentences and zero-based sentence indexes when a label has validated evidence. No extra charge or change to decisions. — e.g. `1` |
| `exclude` | no | Optionally remove confidently labelled engagement bait. Uncertain and unclassified records stay; removed IDs and an owner-bound recall reference are returned. — e.g. `engagement_bait` |
| `fit` | no | Use goal to replace confidently irrelevant records with recall stubs. Keeps first, last, relevant and uncertain records; preserves order. Full evidence is stored for 24 hours when available. — e.g. `goal` |
| `goal` | no | Required with fit=goal: describe the task in at most 600 characters. Supplied text is task context, never permission to act on posts. — e.g. `Find concrete product complaints` |
| `tokens` | no | Optional soft response target for fit=goal, 256–32000. Safety and evidence retention take priority, so the returned result may exceed it. — e.g. `4000` |
| `limit` | no | Optional requested distinct results, 1–120. Collection stops at the existing request spending and time limits, so fewer may be returned. Inspect partial, window_stop_reason and pages_fetched. Keep query, region, sort, dates and limit unchanged with the opaque cursor. Omit for one native page. — e.g. `120` |

**Cost:** 4 credits at list price per successful uncached response; variable-cost operations may quote or settle a different charge. Confirmed uncharged or refunded failures report zero. Pending reconciliation can report credits_used:null; retain request_id and the original idempotency key.

```bash
curl "https://www.monocrawl.com/v1/tiktok/search?query=coffee%20brewing" \
  -H "x-api-key: mn_your_key_here"
```

### Response data

Illustrative abbreviated data; documented core fields are optional and may be null. Additional source fields are allowed.

```json
{
  "items": [
    {
      "id": "7350000000000000002",
      "url": "https://www.tiktok.com/@example_creator/video/7350000000000000002",
      "caption": "A desk setup",
      "created_at": "2026-09-01T12:00:00.000Z",
      "duration_s": 15,
      "stats": {
        "plays": 1200,
        "likes": 100,
        "comments": 12,
        "shares": null,
        "saves": 8
      }
    }
  ],
  "cursor": null,
  "has_more": null
}
```

[Field reference](https://www.monocrawl.com/docs/endpoints/tiktok/search#response) · [JSON Schema for response.data](https://www.monocrawl.com/schemas/tiktok/search.json)