Score hooks. Predict virality.
Without leaving Claude.
Hooklayer is the viral-content intelligence MCP for Claude power users. 9 tools that test hooks, predict virality, find templates, and analyze creators , inside your AI workspace, not another SaaS tab. Test the angle before you waste a day filming. 100K+ viral TikTok, Instagram Reels, and YouTube Shorts indexed.
50 free credits · no card · 30 seconds to first call
Hooklayer is a Model Context Protocol (MCP) server that exposes 100,000+ analyzed viral TikTok, Instagram Reels, and YouTube videos as 9 agent-callable tools organized into 4 workflow layers: Score (hook + virality scoring), Research (trends + templates + video search), Analyze (creator account + voice extraction), and Generate (URL-to-script remix).
Unlike content-generation tools (Arcads, Creatify, Higgsfield, Predis), Hooklayer is positioned for content testing, score the hook before you shoot, predict the virality before you film, find proven templates before you write. Per Kit's April 2026 creator survey (n=550), 82.7% of creators report wanting AI ideation/research partners vs 39.3% who want email drafting. Hooklayer meets them inside Claude (69.8% creator adoption per the same survey) instead of asking them to open another dashboard tab.
Listed in 6 MCP directories: Anthropic Official MCP Registry, Cursor Directory, Smithery (98/100 quality score), Glama, MCP.so, and punkpeye/awesome-mcp-servers. Free tier: 50 credits, no card. Median hook-scoring latency: <400ms. Works with Claude Desktop, Cursor, n8n MCP node, and any MCP 2024-11-05 or 2025-06-18 compliant client.
See analyze_account auto-fire the next 3 tool calls.
3 minutes. One TikTok handle goes in. Claude reads the recommended_chain field and fires match_voice, trend_pulse, and viral_remix on its own. No prompt engineering.
Prefer to try it? The playground runs the full chain on any TikTok handle, no signup, no card.
One tool. The whole agentic chain.
analyze_account takes one TikTok handle and returns deep intelligence plus a recommended_chain field , pre-filled next-tool calls your agent fires automatically. One call in, four calls out.
POST /v1/account/analyze
{
"handle": "@mrbeast",
"platform": "tiktok"
}High-signal voice DNA, consistent across top 5 videos
Verify their formula maps to current trends in challenge_videos
Their #2 video has the highest copyable structure
Claude reads recommended_chain and calls match_voice → trend_pulse → viral_remix on its own. No prompt engineering.
Stop shipping AI slop.
Every n8n / Claude / Cursor content workflow has the same problem: generic LLM calls produce hooks that start with "Hey guys! Today I want to talk about..." and get suppressed by every algorithm.
Hooklayer sits between the LLM and the publish step. Score every hook against proven viral patterns. Reject anything below 70. Auto-rewrite to 85+. One MCP call, one credit.
curl -X POST https://hooklayer.dev/v1/hook/score \
-H "Authorization: Bearer hl_live_..." \
-H "Content-Type: application/json" \
-d '{
"text": "3 things I wish I knew",
"platform": "tiktok"
}'Not just Claude power users.
Hooklayer scores hooks, predicts virality, and extracts Viral DNA the same way for every user, but the workflow shape changes by role. Three audience-specific guides below.
Score every hook before you brief the shoot.
DTC creative teams use Hooklayer to score 10 hook variants in 30 seconds, analyze 5 competitors in 2 minutes, and remix viral URLs into briefable scripts. Cost per test: ~$0.02.
Analyze 50 competitors. Ship 30 ads a week.
Agency tier ($499/mo) pools 32,000 credits across your team for 5-15 client rosters. Weekly competitor sweeps + batch hook testing + client-specific script generation, one flat rate.
Deep viral research with rubrics, not vibes.
Every score returns 6 named signals with evidence quotes suitable for citation. 8 attack vectors documented. Provenance blocks on every response. Rubric-based, not black-box.
9 tools. 4 layers. One Bearer token.
Score → Research → Analyze → Generate. Every layer exposed as both REST endpoints and MCP tools, so Claude Desktop, Cursor, and n8n hit the same surface your scripts do.
Score
Test before you post.POST /v1/hook/scoreScore a hook 0–100 against proven viral patterns. Returns 3 stronger rewrites + matched pattern + percentile.
POST /v1/predictScore a draft script for viral potential. Adversarial attack-vector check + SHIP / REWORK / NO-GO verdict.
Research
Find what is already working.GET /v1/trendsRising opportunities + saturated patterns per niche. Format / hook / style trends, not news topics.
GET /v1/templatesNiche-fit ranked templates with hook patterns, format structures, and example URLs. Optional query narrows to a specific angle.
GET /v1/searchKeyword search across 100K+ viral videos. Region + window + min-views filters. Auto-retries when defaults return empty.
Analyze
Read creators and voices.POST /v1/account/analyzeTikTok handle → full Viral DNA (scores, format fingerprint, steal map) + recommended_chain pre-filling your next 3 tool calls.
POST /v1/voice/matchExtract a creator's voice signature from reference samples. Returns DNA metrics + voice fingerprint your agent reuses across drafts.
Generate
Produce in-voice scripts.POST /v1/remixURL or transcript → fresh script mirroring the original viral DNA. Scene structure, energy pattern, camera shots, native CTA rotation.
Drop into Claude Desktop in 30 seconds.
Hooklayer ships as a hosted MCP server. No local install, no stdio, no compile step. Add the URL + your Bearer token to your client config and every tool is callable from your agent runtime.
{
"mcpServers": {
"hooklayer": {
"url": "https://hooklayer.dev/api/mcp",
"transport": "http",
"headers": {
"Authorization": "Bearer hl_live_..."
}
}
}
}Stop shipping AI slop.
50 free credits. No card. Live in your agent in under 5 minutes.