# Kuaishou API in Python: videos, accounts and comments (2026)

> Read public Kuaishou data from Python with requests: video search with filters, video detail, an account's videos, comments and replies. The 3x… IDs, paging, full code and cost.

Canonical page: https://okdata.online/guides/kuaishou-api-python/

Five endpoints, in the order a script uses them: Kuaishou video search, video detail, account videos, video comments and comment replies, each a plain GET with an API key and no Kuaishou login. The worked example below is 5 calls and costs $0.046; a failed call is not charged.

Updated 2026-11-12

Kuaishou (快手) data from Python usually starts with an open-source crawler. MediaCrawler, the best known, supports Kuaishou search, video IDs, comments and creator pages, and works by logging in through a Playwright browser with a QR code scanned in the app, then reusing the saved login state. That is a sound choice if you run it often and keep the browser and account healthy. This guide is the other option: each call is an HTTPS GET with an API key, no Kuaishou account, cookie, QR login or signature code, and `requests` as the only dependency.

You end with two CSV files: one of videos with plays, likes, comments, shares and saves, and one of comments and replies with text, likes, time and IP location.

## Step by step

Set the key once: `export OKDATA_API_KEY=okd_...`. Every answer has `code` (`"ok"` on success); lists come in `data.items` with a `page`, single objects in `data.item`.

Three kinds of ID matter on Kuaishou. A video's ID is the `3x…` part of its web link, as in `https://www.kuaishou.com/short-video/3xm4f8kq2ze9wbs`. An account's ID is a number such as `2318476509`; the 快手号 handle people put in their bio is something else and shows up as `author.username`. A comment's ID is also a number.

### Step 1: search videos

[Kuaishou video search](https://okdata.online/apis/kuaishou/videos/search/) takes `query` plus three optional filters: `sort` (`relevance` by default, `latest` or `popular`), `published_within` (`day`, `week` or `month`) and `duration` (`under_1_min`, `1_to_5_min` or `over_5_min`). One call returns about 10 videos for $0.02, each with the caption in `text`, the author, and `metrics.views` (plays), `metrics.likes`, `metrics.comments`, `metrics.shares` and `metrics.favorites`.

```bash
curl -G "https://api.okdata.online/v1/kuaishou/videos/search" -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'query=露营装备' --data-urlencode 'sort=popular' \
  --data-urlencode 'published_within=month' --data-urlencode 'duration=under_1_min'
```

### Step 2: refresh one video

[Kuaishou video detail](https://okdata.online/apis/kuaishou/videos/get/) takes `id` (the `3x…` ID) or `url`, and a `v.kuaishou.com` short link from the share button works as `url`. At $0.002 it is the cheap way to re-read a video's counts later, for example once a day to track growth.

```bash
curl -G "https://api.okdata.online/v1/kuaishou/videos/get" -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'id=3xm4f8kq2ze9wbs'
```

### Step 3: list the author's videos

[Kuaishou account videos](https://okdata.online/apis/kuaishou/users/videos/list/) takes the numeric user ID as `id`, which you read from a video's `author.id`. `sort` is `latest` (the default) or `popular`. About 20 videos come back per page for $0.02.

```bash
curl -G "https://api.okdata.online/v1/kuaishou/users/videos/list" -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'id=2318476509' --data-urlencode 'sort=popular'
```

### Step 4: read the comments

[Kuaishou video comments](https://okdata.online/apis/kuaishou/videos/comments/list/) takes the video's `id` or `url` and returns top-level comments newest first, about 10 a page, for $0.002. Each has `id`, `text`, `author.name`, `likes`, `published_at` and `ip_location`. The next page comes from sending `page.next_cursor` back as `cursor`.

```bash
curl -G "https://api.okdata.online/v1/kuaishou/videos/comments/list" -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'id=3xm4f8kq2ze9wbs'
```

### Step 5: open a reply thread

[Kuaishou comment replies](https://okdata.online/apis/kuaishou/comments/replies/list/) needs the comment's ID as `id` and its video's `3x…` ID as `post_id`. About 20 replies come back per page for $0.002, each with `parent_id` pointing at the comment.

```bash
curl -G "https://api.okdata.online/v1/kuaishou/comments/replies/list" -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'id=927364810552' --data-urlencode 'post_id=3xm4f8kq2ze9wbs'
```

### Step 6: the full script

One function makes every call: it checks `body["code"]`, returns `None` for deleted videos, and retries rate limits and busy upstreams with the same `Idempotency-Key`, so a retry is never charged twice. A generator handles paging with a page cap.

```python
import csv
import os
import time
import uuid

import requests

API = "https://api.okdata.online"
KEY = os.environ["OKDATA_API_KEY"]
QUERY = "露营装备"
VIDEO_PAGES = 1     # pages of the author's videos; each page is one call
COMMENT_PAGES = 1   # pages of comments; each page is one call


def call(path, **params):
    headers = {"Authorization": f"Bearer {KEY}", "Idempotency-Key": str(uuid.uuid4())}
    for attempt in range(4):
        body = requests.get(API + path, headers=headers, params=params, timeout=60).json()
        if body["code"] == "ok":
            return body
        if body["code"] == "not_found":
            return None  # deleted or private: not charged
        if body.get("retry") in ("yes", "same_key"):
            time.sleep(body.get("retry_after_seconds") or 2 ** attempt)
            continue
        raise RuntimeError(f"{body['code']}: {body['message']} (request {body['request_id']})")
    raise RuntimeError(f"{path}: still failing after 4 tries")


def paged(path, max_pages, **params):
    cursor = None
    for _ in range(max_pages):
        body = call(path, **params, **({"cursor": cursor} if cursor else {}))
        if body is None:
            return
        yield from body["data"]["items"]
        if not body["page"]["has_more"]:
            return
        cursor = body["page"]["next_cursor"]


def video_row(v):
    m = v["metrics"]
    return [v["id"], v["published_at"], v["author"]["id"], v["author"]["name"], v["text"],
            m["views"], m["likes"], m["comments"], m["shares"], m["favorites"]]


hits = call("/v1/kuaishou/videos/search", query=QUERY, sort="popular", published_within="month")["data"]["items"]
if not hits:
    raise SystemExit("no videos found")
top = max(hits, key=lambda v: v["metrics"]["views"] or 0)

detail = call("/v1/kuaishou/videos/get", id=top["id"])
video = detail["data"]["item"] if detail else top
author_videos = list(paged("/v1/kuaishou/users/videos/list", VIDEO_PAGES, id=video["author"]["id"], sort="popular"))
comments = list(paged("/v1/kuaishou/videos/comments/list", COMMENT_PAGES, id=video["id"]))

replies = []
if comments:
    best = max(comments, key=lambda c: c["likes"] or 0)
    replies = list(paged("/v1/kuaishou/comments/replies/list", 1, id=best["id"], post_id=video["id"]))

with open("kuaishou_videos.csv", "w", newline="", encoding="utf-8-sig") as f:
    out = csv.writer(f)
    out.writerow(["video_id", "published_at", "author_id", "author", "caption", "plays", "likes", "comments", "shares", "saves"])
    out.writerows(video_row(v) for v in [video, *author_videos])

with open("kuaishou_comments.csv", "w", newline="", encoding="utf-8-sig") as f:
    out = csv.writer(f)
    out.writerow(["video_id", "comment_id", "parent_id", "text", "user", "likes", "ip_location", "published_at"])
    for c in comments + replies:
        out.writerow([video["id"], c["id"], c["parent_id"], c["text"], c["author"]["name"], c["likes"], c["ip_location"], c["published_at"]])
```

That is 5 calls, the worked example. The cheap endpoints are detail, comments and replies; search and account videos cost more per call, so cap those first. Fifty comment pages, about 500 comments, are $0.10.

## Good to know

- **Follower counts.** Videos carry the author's ID, name and verification, but `author.followers` is null. [Kuaishou account profile](https://okdata.online/apis/kuaishou/users/get/) has followers, total likes and video count for an account ID.
- **Reply counts.** Kuaishou comments come back with `replies` null, so pick the threads to open by likes or by reading them, then call replies.
- **Order of comments.** Newest first. There is no hottest-first option, which makes the comments endpoint good for checking what people say this week.
- **Photo posts.** Some search results are photo sets, not videos; their `type` is `image`.
- **Extra fields.** With `view=full`, posts add `ext.duration_ms` and `ext.ip_location` (where the post was published from). `fields=` trims the answer. Neither changes the price.
- **Free cases.** Deleted or private videos answer `not_found` and are free, like every other error. Retries with the same `Idempotency-Key` within 24 hours return the stored result with `meta.replayed: true`.

For a daily look at what is trending on Kuaishou, the [Kuaishou hot list](https://okdata.online/tools/hot-search/kuaishou/) is one call. Every Kuaishou endpoint is listed on the [Kuaishou API page](https://okdata.online/apis/kuaishou/).

## What the example costs

One filtered search for 露营装备 (camping gear), the detail of the most-played video, one page of its author's videos, one page of its comments and the replies under the most-liked comment

| Endpoint | Calls | Price | Subtotal |
|---|---|---|---|
| [Kuaishou video search](https://okdata.online/apis/kuaishou/videos/search/) | 1 | $0.02 | $0.02 |
| [Kuaishou video detail](https://okdata.online/apis/kuaishou/videos/get/) | 1 | $0.002 | $0.002 |
| [Kuaishou account videos](https://okdata.online/apis/kuaishou/users/videos/list/) | 1 | $0.02 | $0.02 |
| [Kuaishou video comments](https://okdata.online/apis/kuaishou/videos/comments/list/) | 1 | $0.002 | $0.002 |
| [Kuaishou comment replies](https://okdata.online/apis/kuaishou/comments/replies/list/) | 1 | $0.002 | $0.002 |
| Total | 5 |  | $0.046 |

Prices in force when this page was built. A failed call costs nothing.

## In your AI (MCP)

Connect the MCP server once, then say what you want. The agent finds these endpoints, reads their prices and calls them on your balance.

Say: Search Kuaishou for 露营装备 videos from the last month, sorted by popularity. Take the most-played one, give me its stats, its author's most popular videos, its newest comments and the replies under the most-liked comment.

```json
{
  "endpoint": "kuaishou.videos.search",
  "params": {
    "query": "露营装备"
  },
  "idempotency_key": "<a new UUID>"
}
```

MCP: `https://api.okdata.online/mcp`

## Endpoints in this guide

- [Kuaishou video search](https://okdata.online/apis/kuaishou/videos/search/) `kuaishou.videos.search`: $0.02 per call · about 10 videos per page
- [Kuaishou video detail](https://okdata.online/apis/kuaishou/videos/get/) `kuaishou.videos.get`: $0.002 per call
- [Kuaishou account videos](https://okdata.online/apis/kuaishou/users/videos/list/) `kuaishou.users.videos.list`: $0.02 per call · about 20 videos per page
- [Kuaishou video comments](https://okdata.online/apis/kuaishou/videos/comments/list/) `kuaishou.videos.comments.list`: $0.002 per call · about 10 comments per page
- [Kuaishou comment replies](https://okdata.online/apis/kuaishou/comments/replies/list/) `kuaishou.comments.replies.list`: $0.002 per call · about 20 comments per page

## More about Kuaishou

- [Kuaishou API](https://okdata.online/apis/kuaishou/)
- [What is a Kuaishou video ID, and how do I use a v.kuaishou.com link?](https://okdata.online/answers/kuaishou-video-id-short-link/)
- [Comments API](https://okdata.online/roles/posts-comments-list/)
- [Comment replies API](https://okdata.online/roles/comments-replies-list/)
- [Account posts API](https://okdata.online/roles/accounts-posts-list/)
- [Post detail API](https://okdata.online/roles/posts-get/)
- [Post search API](https://okdata.online/roles/posts-search/)
- [Kuaishou hot list: today’s list](https://okdata.online/tools/hot-search/kuaishou/)

## Sources

- [MediaCrawler README (GitHub)](https://github.com/NanmiCoder/MediaCrawler) (read 2026-11-12)