Agentic Video Editing via MCP

That's not a hypothetical. That's a direct quote from a video editor at a 9-figure beauty brand I just got off the call with. Their team was drowning in testimonial edits - spending 20 minutes per video in CapCut, manually cutting compliance issues, moving hooks around.

Now? They type instructions into a Google Sheet cell. The a/gentic Video Editor Agent edits the video. A new URL appears in the next cell. Done.

No timeline scrubbing. No export waiting. Just natural language commands and instant results.

The Creative Team's New Reality

During our call, their creative director dropped another truth bomb:

Think about that. They have tens of thousands of video assets. Traditional DAM systems turned them into a graveyard. Now every scene is instantly searchable by meaning, mood, audio, or visual elements.

The Breakthrough: Trusting AI to Think

The real innovation wasn't in the editing capabilities - it was in letting go of deterministic programming and trusting AI to reason.

Initially, I tried programming every edge case:

• If sentence ends with period, cut here

• If there's a pause, include extra frames

• If words run together, use alternate logic

After days of if-else-then spaghetti code, I had an epiphany:

Why am I trying to outthink an AI that can already reason better than my code?

The Google Sheets Magic

Here's what makes this system cool - it lives where your team already works:

1. Evaluator agent reviews testimonial and suggests edits in column A

2. Editor simply types refined instructions in column B

3. AI agent processes the edit and returns new video URL in column C

No software app to use. No context switching. Just a spreadsheet that edits videos.

The Technical Pivot That Changed Everything

Instead of complex sentence-parsing algorithms, I:

1. Enhanced our transcription with word-level timestamps

   • Used Open AI’s whisper-1 model with timestamp_granularities to return exact start/end times for every word, set to include both "segment" and "word" levels, which returns exact start/end timestamps for every individual word in the transcript.

2. Built an MCP tool in the dtc.sh AdMaker MCP server that accepts natural language + timecodes

   • The human describes what they want in plain English

   • The AI agent (using o3's reasoning) determines exact cut points

   • The tool executes the edits with surgical precision

Agentic Scene Prep

That creative director’s quote about never finding anything is universal problem. Brands accumulate thousands of videos but can't access the value trapped inside them.

Our Scene Prep Agent changed that:

• Vectorized tens of thousands of videos into searchable scenes

• Semantic search finds clips by meaning, not just tags

• Instant discovery of similar successful ad moments

• No more folder diving in Air, Recharm, Frame.io, or Dropbox

Now when they need "testimonials showing morning routines" or "before/after reveals with excitement," the system returns relevant scenes in seconds.

Real-World Results: 67% Productivity Increase

Before (Manual Editing):

• 20 minutes per video in CapCut

• ~100 ads per week capacity for the team

• 30 hours weekly just on editing

• Constant context switching

• Inconsistent compliance

After (Agentic Editing):

• 3-5 minutes per video via Google Sheets

• 160 ads per week capacity for the team

• Same 30 hours, 67% more output

• Stay in familiar workflow

• Perfect compliance every time

But here's the hidden benefit: cognitive load reduction.

Video editors no longer waste mental energy on repetitive cuts. They focus on creative decisions - which hooks work, what stories resonate, how to maximize emotional impact.

Why This Matters for Every E-commerce Brand

If you're processing creator content at scale, you're facing the same challenges:

• Good hooks buried 30 seconds in

• Compliance landmines everywhere

• Editors drowning in repetitive work

• Assets lost in digital graveyards

• Testing velocity limited by editing time

Agentic video editing solves all of this:

• 10x faster editing through natural language

• 100% compliance with automatic detection

• Instant asset discovery via semantic search

• Unlimited testing with rapid iterations

• Team sanity preserved for strategic work

The Philosophical Shift

Building this system taught me something profound about the future of work:

Old way: Humans do repetitive tasks while AI assists New way: AI handles repetitive tasks while humans strategize

We're not replacing video editors. We're giving them superpowers. That editor who went from 30 to 50 ads per week? They're not working harder. They're working at a higher level.

The Integration That Changes Everything

The Google Sheets integration was the unlock. We could have built a fancy interface, required training, created another tool to learn.

Instead, we met teams where they already work. A spreadsheet. The most boring, familiar interface possible. But that's exactly why it works.

Your media buyers already live in sheets. Your creative teams already share docs. Now those same tools edit videos with AI precision.

What's Next

This video editing agent is joining our suite of e-commerce AI agents:

• Testimonial Evaluator Agent: Scores videos for conversion potential

• Scene Prep Agent: Vectorizes content for semantic search

• Compliance Agent: Ensures regulatory adherence

• Performance Predictor: Estimates ROAS before spending (Coming soon)

Together, they're transforming how brands process creator content at scale.

Start Small, Scale Fast

You don't need to revolutionize everything overnight. Start with one workflow:

1. Pick your highest-volume repetitive task (probably testimonial editing)

2. Implement agentic editing for just that workflow

3. Measure the time savings and quality improvement

4. Expand to other content types

Most brands see positive ROI within the first week. Because when you go from 20 minutes to 3 minutes per edit, the math is undeniable.

Want access to a/gentic’s Video Editor Agent? DM Bora on LinkedIn