Sign Up
Social Media Scraping API

Social media data, built for AI agents

Give Claude, Cursor, OpenAI Assistants, or any agent framework first-class access to YouTube, TikTok, Instagram, and Facebook - transcripts, stats, comments, downloads, search. No scrapers, no proxies, no headless browsers.

  • 33 MCP tools + 47 REST endpoints - same data, two delivery shapes
  • Works with Claude Desktop, Cursor, OpenAI, LangChain & any agent framework
  • TikTok, Instagram, YouTube, Shorts, Facebook, Twitter, LinkedIn covered out of the box
  • Stop maintaining scrapers - managed API absorbs anti-bot + schema churn
No credit card required.

Loved by 20,000+ data lovers

https://api.socialkit.dev/tiktok/summarize?url=tiktok.com/@thepeteffect/video/7522711492140059912
AI Summary
+transcript

Video discusses upcoming social media trends including AI-generated content...

• Full transcript available
Engagement
+15 metrics
94%
Rate
8.2
Score
• Views, likes, comments...
Topics
+keywords
AI ContentTrendsMarketing...
• Sentiment, entities...
Quick start

Connect your agent in one line

Install the skill in any coding agent, or drop the MCP config into your client. Same access key for both - no scrapers, no wrapper code.

Agent skill — Claude Code, Cursor, Codex
npx skills add SocialKit-API/skills --all
MCP server — Claude Desktop & MCP clients
{
  "mcpServers": {
    "socialkit": {
      "url": "https://mcp.socialkit.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_KEY"
      }
    }
  }
}

Your agents already know how to use this

A first-party Claude Code skill, a hosted MCP server, and a plain REST API - all official, all maintained.

Agent skill

A drop-in skill for Claude Code, Cursor, and Codex. It teaches them every endpoint, param, and response shape.

terminal
$ npx skills add \
    SocialKit-API/skills --all
  ✓ added 7 SocialKit skills

> summarize the last 5 reels from
  @nasa and rank them by views
Read the docs

MCP server

One config block. Claude, Cursor, or any MCP client gets every SocialKit endpoint as a first-class tool.

mcp config
{
  "mcpServers": {
    "socialkit": {
      "url": "https://mcp.socialkit.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_KEY"
      }
    }
  }
}
Read the docs

REST API

Pipe-friendly and framework-agnostic. Wrap the endpoints as tools in OpenAI, LangChain, CrewAI, or anything with function-calling.

terminal
$ curl https://api.socialkit.dev/\
    tiktok/stats?url=... \
    -H "x-access-key: YOUR_KEY"

  → { likes, views, shares, ... }
    1 credit used
Read the docs

How agents use SocialKit

From zero to structured social data in four steps. The same access key works across the skill, MCP, and the REST API.

01

Grab an access key

One free key from the dashboard authenticates every request across the REST API, MCP, and the skill. 20 credits to start, no card.

02

Install the skill or add MCP

Run one npx command in Claude Code or Cursor, or paste the MCP block into any client. No scrapers, no proxies, no headless browsers.

03

Let it pick the endpoint

The agent reads a social URL, detects the platform, and calls the matching endpoint: transcript, stats, comments, search, or download.

04

Get structured data back

Clean JSON in the same shape across every platform. Pipe it straight into analysis, another agent, or your own app.

Agent recipes

Six things you can build this week

Each one is achievable in a single afternoon with SocialKit + your favorite agent framework.

Competitor monitoring agent

Schedule an agent to poll your competitors' channels daily. New video drops → it pulls the transcript, summarizes, runs sentiment on the top comments, and posts the brief to Slack. ~30 lines of code.

youtube_videos
youtube_summarize
youtube_comments

Content research assistant

"Find me the top 5 TikToks about productivity this week and tell me what hooks they're using." Agent searches, ranks by views, pulls transcripts, identifies pattern. One prompt.

tiktok_search
tiktok_transcript
tiktok_stats

Comment sentiment dashboard

Drop any URL in your AI chat. It detects platform, fetches up to 200 comments, runs sentiment classification, returns a JSON breakdown with the most-positive and most-negative outliers.

instagram_comments
tiktok_comments
youtube_comments

Trend detection

Subscribe an agent to a hashtag. Every hour it polls, ranks new videos by view velocity, and pings you when something is breaking. The agent decides what 'breaking' means based on history.

tiktok_hashtag_search
instagram_reels_search

Audio-first content pipeline

Agent downloads a Reel, transcribes it, generates a blog post from the transcript, and uploads to your CMS. Each step is one tool call. Plug it into any agent framework.

instagram_download
instagram_transcript

Personalized brief on any URL

Paste any social URL into your AI; it figures out which platform it's from, fetches stats and transcript, and writes a 3-line summary in your tone. Universal "what is this?" button.

youtube_stats
tiktok_stats
instagram_stats
facebook_stats
Common tool calls

Small, predictable actions your agent can chain

Every endpoint takes a URL or a query and returns clean JSON in a consistent shape. One agent fetches, another analyzes, a third decides what to do next.

Browse the API reference
  • youtube_transcript: full transcript with per-segment timestamps
  • tiktok_stats: views, likes, shares, and comment counts for a video
  • instagram_comments: paginated comments to score sentiment
  • tiktok_hashtag_search: discover trending videos for a hashtag
  • youtube_channel_stats: subscribers, bio, and recent uploads
  • video_download: grab an MP4 for downstream processing

Why not just scrape?

Agents are non-deterministic callers. Scrapers assume the opposite — that's why they keep breaking.

Approach
Setup
Maintenance
Agent fit
Puppeteer / Playwright
Days
Weekly breakage
Painful
yt-dlp + duct tape
Hours
Frequent
Glue-heavy
RapidAPI wrappers
Minutes
Listings vanish
OK for prototypes
SocialKit REST API
5 minutes
None — managed
Standard tool
SocialKit MCP
30 seconds
None — managed
First-class

Fast and consistent

Same response shape across all platforms. 1-3s on cache miss, sub-second on cache hit. Failed calls are free.

All the data, no scraping

Transcripts, stats, comments, downloads, search, channel data across YouTube, TikTok, Instagram, and Facebook. We handle anti-bot, proxies, and platform churn.

Composes with your stack

REST for any framework, MCP for Claude/Cursor, OpenAPI 3.0 spec for code generation. Same key works across all three surfaces.

Stop building social-media scrapers

Hook your agent up in under a minute. Free tier is enough to ship a working prototype.

20 free creditsNo credit card33 MCP tools • 47 REST endpoints

Frequently Asked Questions

Ready to Extract Social Media Videos Data?

Extract video summaries, transcripts, comments, and engagement metrics from social media content with our powerful Social Media Data Extraction API.

Start Extracting for FreeNo credit card required.
https://api.socialkit.dev/youtube/summarize?url=youtube.com/watch?v=dQw4w9WgXcQ
AI Summary
+transcript

Video discusses upcoming social media trends including AI-generated content...

• Full transcript available
Engagement
+15 metrics
94%
Rate
8.2
Score
• Views, likes, comments...
Topics
+keywords
AI ContentTrendsMarketing...
• Sentiment, entities...