Automate YouTube Shorts creation with Ssemble AI Clipping MCP Server + n8n

Built with n8n
1

Hey n8n community :waving_hand:

I wanted to share how you can use n8n’s built-in MCP Client Tool to connect to the Ssemble AI Clipping MCP server and automate short-form video creation from YouTube videos — no custom nodes needed.

:rocket: What is Ssemble AI Clipping?

Ssemble is a video processing platform that takes YouTube videos and automatically creates TikTok/Reels/Shorts-style clips with:

  • AI-selected highlight moments with viral scoring
  • 9+ caption templates (Karaoke, Bold, Minimal, etc.)
  • 45+ background music tracks
  • 30+ gameplay overlays (Minecraft, GTA, Subway Surfer, etc.)
  • 159+ meme hook clips (2–5 second attention grabbers)
  • AI-generated titles, descriptions, and hashtags

We recently published an MCP server that exposes all of this as 9 tools any MCP client can use.

:electric_plug: How to connect it in n8n

Since n8n has native MCP support via the MCP Client Tool node, setup is straightforward:

Step 1 — Prerequisites

  • Self-hosted n8n (v1.88.0+)
  • Set environment variable:
    N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true
  • Get a Ssemble API key from: https://app.ssemble.com (Settings → API Keys)

Step 2 — Add MCP Client Tool node

  • Add MCP Client Tool node

  • Transport: SSE

  • URL: https://mcp.ssemble.com/mcp

  • Authentication: Header Auth

    • Header Name: X-Ssemble-API-Key
    • Header Value: sk_ssemble_your_key_here

  • Tools: All (or select specific)

Step 3 — Connect to AI Agent

  • Add an AI Agent node (Claude, OpenAI, etc.)
  • Connect MCP Client Tool as a tool input
  • The agent now has access to all Ssemble tools

:toolbox: Available tools (9 total)

Tool Description
create_short Submit YouTube URL for clipping (returns request ID)
get_status Check progress (0–100%)
get_shorts Get clips + download URLs + viral scores
list_requests List all requests
delete_request Delete a request
list_templates Browse caption styles
list_music Browse music tracks
list_game_videos Browse gameplay overlays
list_meme_hooks Browse meme clips

:light_bulb: Example workflows

1. Content repurposing pipeline

RSS → detect new YouTube video → create_short → wait/poll → get_shorts → post to Slack/Discord

2. Batch processing

Google Sheet (URLs) → loop → create_short → collect request IDs → poll → save results back

3. AI-powered automation

Webhook → user input (“make it funny with gaming overlay”) → AI Agent chooses tools → generate clips → return results

:hourglass_not_done: Important: Async processing pattern

Video processing takes 5–30 minutes, so:

  1. create_short → returns requestId instantly
  2. get_status → check progress
  3. get_shorts → fetch results

:backhand_index_pointing_right: Tip: Add a Wait node (5–10 min) or loop with polling until completed.

:link: Links

Would love to see what you build with this :rocket:
Happy to answer any questions!

5

Hey friend!

Cool setup — this is a great use case for n8n.

For anyone looking for an alternative that works with both MCP and direct HTTP Request nodes, I’ve been using WayinVideo API (https://wayin.ai/api) for the same kind of workflow.

A few things that work well for me:

  • AI Clipping — automatically picks the best viral moments, adds captions, reframes to 9:16 — all in one API call

  • Find Moments — you can describe what you want in plain English like “product demo” or “funny reaction” and it finds those exact clips

  • No MCP required — works directly with n8n’s HTTP Request node, so it runs on both Cloud and self-hosted without any extra setup

  • Processing time is around 90 seconds for most videos — much faster than 5–30 minutes

The flow is simple:

Form Trigger / Webhook → POST video URL to WayinVideo API → Wait 90s → GET clips → each clip has download link, viral score, title, tags

Also supports YouTube, Zoom, TikTok, Vimeo, and more — just paste the URL.

Docs here if anyone wants to check it out: https://wayin.ai/api-docs/ai-clipping/

Happy to share the n8n workflow JSON too!

1 Like
6

Hey ssemble!

Nice guide — the webhook callback tip from Benjamin is a good catch, polling with Wait nodes does eat up execution slots fast.

Been using a similar setup for YouTube Shorts creation but without the MCP layer. The one thing that bugged me about MCP-based tools is they only work on self-hosted n8n with that environment variable set — which cuts out anyone on n8n Cloud.

What I switched to uses plain HTTP Request nodes so it runs anywhere — Cloud or self-hosted, no extra config needed. You just POST the YouTube URL, get back clips ranked by viral score with titles, timestamps, and download links. It also does 9:16 auto-reframe and animated captions in the same call, so no separate editing step.

The big difference I noticed is processing time — most videos come back in around 90 seconds instead of 5–30 minutes, which makes the polling pattern much simpler to manage.

Also has a Find Moments feature where you describe what you want in plain English — like “product demo” or “audience laughing” — and it pulls exactly those clips. Useful when you don’t just want the top viral moments but something specific.

Docs here if anyone wants to compare: https://wayin.ai/api-docs/ai-clipping/

Happy to share the workflow JSON!

7

Thanks for sharing. This guide really helps me.

8

Can you explain how to use n8n’s built-in MCP Client Tool to connect to the Ssemble AI Clipping MCP server for automatically turning YouTube videos into short-form clips, and walk me through the setup steps, available tools, and example workflows?

9

the mcp client tool route is the right call here over building a custom http node.

having the model negotiate the clipping boundaries based on audio transcripts avoids so much messy manual timestamp math.

how are you handling render queues when multiple long videos arrive back-to-back? do you throttle executions in n8n or queue them upstream?