Jonathon's AI Wiki

ScrapeCreators — Social Media Scraping API for AI Pipelines

8 min read

Source: Scrapecreators Landing Page (2026-05-08)

ScrapeCreators is a social-media scraping API by Adrian Horning (Austin, TX) — one x-api-key header against 20+ platforms (TikTok, Instagram, YouTube, Facebook, X, Reddit, LinkedIn, Pinterest, Threads, Bluesky, Truth Social, Twitch, Spotify, SoundCloud, Snapchat, Kick, plus link-in-bio platforms and ad libraries from Google / Facebook / TikTok / LinkedIn). 100 free credits no-card; pay-as-you-go from 497 / 500k credits to enterprise. No rate limits, no SDK required, JSON responses at ~3.1s avg. Ships an official MCP server (@scrape-creators/mcp), CLI tool, and a first-party Claude Code skill. Sits alongside TinyFish in the “vendor data infrastructure for AI agents” category but goes deeper on social-platform-specific endpoints — TikTok alone has 20 endpoints including audience demographics and live-comment scraping. The Karpathy last30days skill’s SessionStart hook calls it out by name as the gap-filler for Reddit comments + TikTok + Instagram.

Key Takeaways

  • 20+ platforms behind one API key. Nine “social” platforms with multiple endpoints each (TikTok 20, Instagram 12, YouTube 12, Facebook 9, X/Twitter 6, LinkedIn 4, Reddit 5, Pinterest 4, Threads 5), five link-in-bio platforms (Linktree, Komi, Pillar, Linkbio, Linkme), and 16 additional platforms covering TikTok Shop, Google Search, Twitch, Spotify, SoundCloud, Kick, Snapchat, Bluesky, Truth Social, Amazon Shop, plus age-and-gender detection and four ad libraries (Google, Facebook, TikTok, LinkedIn). Single x-api-key header for all of them.
  • Free tier is 100 credits, not 10,000. The Karpathy project’s last30days SessionStart hook quotes “10,000 free API calls, no credit card” but the live landing page (fetched 2026-05-08) shows 100 credits as the free-tier figure. Either a stale promo or referral-code-gated tier; treat 100 as the canonical public figure. ^[ambiguous]
  • Pay-as-you-go pricing scales linearly with credits. 1.88 / 1k requests). 0.99 / 1k requests, ~50% cheaper at scale). Enterprise custom. Credits never expire — no usage-or-lose, no monthly subscription pressure. 1 request = 1 credit for most endpoints.
  • No rate limits, no SDKs, raw HTTP. “Unlimited concurrent requests, no per-minute caps or throttling.” curl works. Auth is x-api-key: YOUR_API_KEY. JSON responses. The vendor’s pitch is explicitly against the SDK abstraction layer — keep it as raw as a curl command. Same shape as the Meta Ads CLI / ElevenLabs voice agents / Railway Remote MCP vendor-direct pattern.
  • Three first-party AI integration paths.
    • MCP server at @scrape-creators/mcp — Claude Desktop, Cursor, any MCP-compatible client.
    • CLI tool — “pipe-friendly, cron-friendly, CI-friendly” for shell scripts and batch operations.
    • Claude Code skill — first-party skill teaching Claude every endpoint and parameter. Drop-in integration. Choosing between them maps to the Skills decision framework: MCP for live agentic use, CLI for batch / cron / CI, Skill for narrative-driven Claude Code sessions.
  • Real production scale claims. 1M+ requests/day, 98.2% success rate, 3.12s average response time. These are vendor-supplied — but plausible given the testimonials from named founders (Eric Nowoslawski / Growth Engine X, Gil / subscribr.ai, Cameron Blackwood, others) and the YC-startup-customer angle.
  • Six headline use cases.
    1. Clipper payouts — pull views/likes/shares to automate creator payouts (TikTok-Shop-affiliate / influencer-pay-per-view economy).
    2. Ad spy & intelligence — Facebook Ad Library, TikTok Ads, Google Ads at scale. Closest competitor analog: ad-intel platforms like Foreplay, Pipi, Atria.
    3. Analytics dashboards — live engagement, follower growth, creator metrics.
    4. Link-in-bio tools — auto-fill profiles, bios, recent posts during signup.
    5. Video intelligence — transcripts, comments, metadata for AI pipelines (RAG-the-creator-economy).
    6. Personal & brand tracking — solo creators, agencies, brands monitoring categories.
  • TikTok depth is the moat. The 20 TikTok endpoints include User Audience Demographics (rare in public APIs), TikTok Live, Comment Replies, Profile Region, and Search Users. This is a stronger TikTok surface than what’s available via the last30days skill’s yt-dlp-based pipeline (which can only get TikTok video metadata, not demographics or live data) or the birdclaw-style cookie-auth approach (Twitter-only). For TikTok-heavy creator-economy workflows, it’s currently the cleanest shortcut.
  • Ad-library access is the second moat. Facebook Ad Library (Meta), TikTok Ads, Google Ad Library, LinkedIn Ad Library — all four major ad libraries surfaced through one API. Enables Meta Ads CLI-style workflows but at the competitive intelligence layer (what are competitors running?) rather than the campaign-management layer (what are you running?).
  • The vendor-relationship pitch. “1-on-1 support from founder Adrian Horning. Fast replies, no ticket queue.” Solo-founder direct-line model — recurring shape across small-but-shipping API vendors (compare to TinyFish’s ICONIQ-Series-A scale, or the Meta Ads CLI solo-author model). Trade-off: lower bus-factor, faster bug fixes, no enterprise contracts.

Where this fits in the wiki

  • agents-agentic-systems/ is the right home — same shape as TinyFish (web infra APIs) and Adaline (LLMOps). All three are “vendor data infrastructure for AI agents” — different domains (social vs web vs LLM ops), same role.
  • Direct competitor coverage in the wiki.
    • TinyFish — web Search/Fetch/Browser/Agent. Overlapping on YouTube and Reddit; ScrapeCreators is broader on social platforms specifically.
    • Apify (Cowork recipe) — generalist scraping marketplace. Different model (per-actor pricing on Apify vs single-key API on ScrapeCreators).
    • last30days skill — research aggregator that already pulls from Reddit/X/YouTube/HN/TikTok/Polymarket/GitHub via mostly-OSS scrapers (yt-dlp etc.). ScrapeCreators is the paid alternative when those scrapers break or hit rate limits, particularly for TikTok/Instagram demographic data.
    • birdclaw — local cookie-auth Twitter archive workspace. Different category (local archive vs API).
  • Vendor-direct tool-calls pattern. Adds a sixth article to the candidate “vendor-direct tool calls” connection (currently 5 articles converging on direct vendor APIs over n8n/Zapier middleware: voice-agents-elevenlabs, cowork-live-artifacts, higgsfield-mcp, meta-ads-cli, railway-remote-mcp). With ScrapeCreators making it 6, this connection is now overdue — flagged in MEMORY.md as project_vendor_direct_tool_calls_pattern.md. No connection article yet exists; would live at wiki/connections/vendor-direct-tool-calls-pattern.md.
  • Reddit/social marketing touch points. Connects to Reddit Marketing (Reddit data for AI-citation tracking), social-media-skills (need fresh post/comment data to score against), and AI Marketing (ad-library-driven competitive intelligence).

Implementation

  • Tool/Service: ScrapeCreators — social-media scraping API across 20+ platforms.
  • Setup:
    • Sign up at https://scrapecreators.com → 100 free credits, no card.
    • Get x-api-key from dashboard.
    • Read docs at https://docs.scrapecreators.com/ (vendor calls them out as “super clear also for very non-technical people”).
    • For Claude integration, install the MCP server (@scrape-creators/mcp) or the Claude Code skill.
  • Cost: Free 100 credits → 1.88 / 1k) → 0.99 / 1k) → enterprise. 1 request = 1 credit. Credits never expire.
  • Integration notes:
    • One x-api-key header for all 30+ APIs.
    • No rate limits / no concurrency caps.
    • No SDK required; raw HTTP works.
    • JSON only.
    • Real-time data — no cached results, so freshness is “what the public timeline shows right now.”
    • For agentic use, prefer MCP server (live tool-calling) or Claude Code skill (narrative).
    • For batch/scheduled use, prefer the CLI (“cron-friendly”).
  • Example call (vendor-supplied): 
    curl "https://api.scrapecreators.com/v1/tiktok/profile?handle=iamsydneythomas" \
  -H "x-api-key: YOUR_API_KEY"

Open Questions

  • Free-tier discrepancy. Karpathy SessionStart hook says “10,000 free API calls;” landing page says “100 credits.” Is 10,000 a referral-code-gated promo, an old tier, or the hook copy is stale? ^[ambiguous]
  • Compliance posture. Public-data scraping is legally fraught (Meta v. BrightData, hiQ v. LinkedIn). What’s ScrapeCreators’ position on platform ToS? Not visible on the landing page — relevant for any production deployment.
  • TikTok Shop endpoint depth. Listed as “additional” platform but no endpoint count. For affiliate/clipper-economy workflows, TikTok Shop product/order data could be the highest-value endpoint; needs docs check.
  • Rate-limit absence. “No rate limits” is unusual for a public scraping API. Either upstream platforms are doing the throttling and ScrapeCreators is a thin pass-through, or they’re absorbing the cost themselves. The 98.2% success rate suggests upstream-driven failures rather than ScrapeCreators-side limits — but production planning should still expect 401/429s downstream.
  • MCP server stability. @scrape-creators/mcp package — version, last release, Node version requirements not documented on the landing page. Worth a npm view before depending on it in production.
  • Credit cost per endpoint. “1 credit = 1 request for most endpoints” — but which endpoints aren’t 1:1? TikTok Live and User Audience Demographics seem like obvious candidates for higher-cost endpoints. Docs would specify.

Try It

  1. Sign up for the free 100 credits. No credit card required. https://scrapecreators.com → email → API key.
  2. Try the curl example. Paste the TikTok profile call from the landing page (/v1/tiktok/profile?handle=...) with your key. Confirm you get JSON back, see what fields are populated.
  3. Install the Claude Code skill. The first-party skill teaches Claude every endpoint with parameter schemas — significantly cheaper than burning credits to discover the endpoint shape.
  4. Pair with last30days. The Karpathy last30days skill’s SessionStart hook recommends ScrapeCreators for Reddit comments + TikTok + Instagram — the platforms its yt-dlp / public-API approach is weakest on. Wire ScrapeCreators in as the supplementary data source for those three; keep yt-dlp etc. for YouTube/HN/etc.
  5. Try one ad-library call. Hit the Facebook Ad Library or TikTok Ads endpoint for a competitor brand. Compare against what you’d see manually in the public Ad Library UI — useful baseline for AI marketing competitive intel work.
  6. For TikTok-heavy creator workflows, burn the demo credits on User Audience Demographics for a single creator. That’s the endpoint hardest to get elsewhere; if the data quality holds up, it’s the load-bearing reason to upgrade to a paid tier.

Graph View

Backlinks