keyapi-tiktok-content-analysis（keyapi 抖音内容分析）

keyapi-tiktok-content-analysis

Analyze TikTok content at scale — extract insights from videos, hashtags, music, and live streams to inform content strategy, trend discovery, and audience engagement research.

This skill provides comprehensive content intelligence on TikTok using the KeyAPI MCP service. It enables deep analysis of video performance, comment sentiment, caption extraction, hashtag reach, music popularity, and live-stream activity — all backed by real-time data and historical analytics datasets.

Use this skill when you need to:

• Analyze the performance trajectory of specific TikTok videos or content creators
• Research hashtag reach and trending topics for content positioning
• Extract video captions and analyze comment sentiment for audience insights
• Identify high-performing music tracks for content creation
• Monitor live-stream activity and audience engagement patterns
• Build content performance benchmarks for editorial or marketing strategy

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Prerequisites

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

MCP Server Configuration

All tool calls in this skill target the KeyAPI MCP server:

Server URL : https://mcp.keyapi.ai
Auth Header: Authorization: Bearer $KEYAPI_TOKEN

Setup (one-time):

# 1. Install dependencies
npm install

# 2. Set your API token (get one free at https://keyapi.ai/)
export KEYAPI_TOKEN=your_token_here

# 3. List all available tools to verify the connection
node scripts/run.js --list-tools

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Analysis Scenarios

Video Nodes

Hashtag Nodes

Music Nodes

Live Stream Nodes

Cross-Domain Search

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Workflow

Step 1 — Identify the Content Analysis Objective and Select Nodes

Clarify the user's goal and map it to one or more nodes. Common research patterns:

• Video performance analysis: Start with get_video_detail, deepen with video_detail_analytics + video_trends_analytics.
• Comment and sentiment analysis: Use get_video_comments + video_comment_keywords.
• Hashtag research: Use search_hashtags to find relevant tags, then get_hashtag_videos for content within the tag.
• Music trend research: Use search_music to find tracks, cross-reference with trending_music (see keyapi-tiktok-intelligence skill).
• Live stream monitoring: Use search_live_streams to find active streams, then get_live_stream_detail with both room_id and user_id for specifics.
• Commerce content analysis: Use video_products_analytics to correlate video content with product sales data.
• Cross-domain keyword research: Use general_search_analytics for a unified view across products, shops, and creators.

Multiple nodes can be selected for cross-dimensional analysis — for example, analyzing a video's engagement trend while simultaneously extracting its comment keywords to understand what drove viewer interest.

Step 2 — Retrieve API Schema

Before calling any node, inspect its input schema to confirm required parameters and valid values:

node scripts/run.js --schema <tool_name>

# Examples
node scripts/run.js --schema video_list_analytics
node scripts/run.js --schema video_comment_keywords

Step 3 — Call APIs and Cache Results Locally

Execute the required tool calls and persist all responses to the local cache.

Calling a tool (using scripts/run.js):

# Single call — result is cached automatically
node scripts/run.js --tool <tool_name> --params '<json_args>' --pretty

# Fetch all pages at once (auto-pagination)
node scripts/run.js --tool <tool_name> --params '<json_args>' --all-pages

# Force a fresh call, skip cache
node scripts/run.js --tool <tool_name> --params '<json_args>' --no-cache

Example — get video details:

node scripts/run.js --tool get_video_detail \
  --params '{"video_id":"7xxxxxxxxxxxxxxxxx"}' --pretty

Example — extract comment keywords:

node scripts/run.js --tool video_comment_keywords \
  --params '{"video_id":"7xxxxxxxxxxxxxxxxx"}' --pretty

Pagination for analytics endpoints:

All *_analytics endpoints use page_num (1-indexed) and page_size (max 10). run.js injects these automatically if not specified.

--page-num 1  --page-size 10   → first page (default)
--all-pages                    → all pages merged into one result

Cache directory structure:

.keyapi-cache/
├── videos/
│   └── {video_id}/
│       ├── detail.json               # get_video_detail / video_detail_analytics
│       ├── captions.json             # get_video_captions
│       ├── comments.json             # get_video_comments
│       ├── comment_keywords.json     # video_comment_keywords
│       ├── interaction_trends.json   # video_interaction_trends
│       ├── trends_analytics.json     # video_trends_analytics
│       ├── products.json             # video_products_analytics
│       └── download_url.json         # get_video_download_url
├── hashtags/
│   └── {hashtag_id}/
│       └── videos.json               # get_hashtag_videos
├── live_streams/
│   └── {room_id}/
│       └── detail.json               # get_live_stream_detail
├── searches/
│   ├── videos/
│   │   └── {md5_of_query_params}.json    # search_videos, video_list_analytics
│   ├── hashtags/
│   │   └── {keyword}.json                # search_hashtags
│   ├── music/
│   │   └── {keyword}.json                # search_music
│   ├── live_streams/
│   │   └── {keyword}.json                # search_live_streams
│   └── general/
│       └── {md5_of_query_params}.json    # general_search_analytics
├── rankings/
│   └── videos_{params_hash}.json         # video_ranking_analytics
└── images/
    └── covers/
        └── {converted_filename}           # batch_download_cover_images output

Cache-first policy:

Before every API call, check whether a cached result already exists for the given entity and node. If a valid cache file exists, load from disk and skip the API call.

Cover image processing (mandatory):

⚠️ Cover image conversion is required for all analytics responses. After every API call that returns content with images, scan all image URLs in the response. If any URL's host matches echosell-images.tos-ap-southeast-1.volces.com, collect all such URLs and issue a single batch_download_cover_images call. Replace the original URLs in your working dataset with the converted URLs returned. Save converted images to .keyapi-cache/images/covers/.

Example — batch cover image conversion (run.js does this automatically):

node scripts/run.js --tool batch_download_cover_images \
  --params '{
    "cover_urls": "https://echosell-images.tos-ap-southeast-1.volces.com/img/abc.jpg,https://echosell-images.tos-ap-southeast-1.volces.com/img/def.jpg"
  }' --pretty

Step 4 — Synthesize and Report Findings

After collecting all API responses, produce a structured content intelligence report tailored to the analysis type:

For video analysis:

1. Video Overview — Title, creator, publish date, view count, likes, comments, shares.
2. Engagement Deep-Dive — 14-day interaction trend, engagement rate vs. category benchmarks, virality signals.
3. Audience Insights — Top comment themes, keyword frequency, positive/negative sentiment signals.
4. Content-Commerce Link — Products featured, sales generated, conversion indicators.
5. Strategic Takeaways — What made this content perform (hook analysis, format, timing, hashtag leverage).

For hashtag/topic analysis:

1. Hashtag Profile — Total video count, view volume, growth trend.
2. Top Content — Highest-performing videos within the hashtag, common formats.
3. Creator Patterns — Who is driving the conversation, creator tier distribution.
4. Opportunity Assessment — Saturation level, engagement quality, best entry timing.

For music analysis:

1. Track Overview — Title, artist, total usage count.
2. Usage Trends — Growth in video adoption over time.
3. Content Fit — Types of content using this track, associated hashtags and themes.

For live stream analysis:

1. Stream Profile — Host, viewer count, products featured, duration.
2. Commerce Performance — GMV generated, conversion signals.
3. Engagement Quality — Peak concurrent viewers, comment activity.

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Common Rules

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Error Handling