Skillget

capcut-ai-editor

mcp
Security Audit
Warn
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 8 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
This MCP server automates "talking head" video editing by integrating with CapCut projects. It reads locally generated subtitles to identify and remove silences and duplicate takes.

Security Assessment
Overall risk: Low. The code scan of 12 files found no dangerous patterns or hardcoded secrets, and the tool does not request dangerous permissions. It operates entirely as a local heuristic utility without requiring API keys. However, users can optionally provide an OpenAI API key via environment variables to enable GPT-enhanced duplicate detection, which will trigger outbound network requests. The tool reads and modifies local CapCut project files directly, so no sensitive system data or remote CapCut credentials are exposed. Just be aware that the edits happen in-place without creating automatic backups.

Quality Assessment
The project is very new and has low community visibility, currently sitting at only 8 GitHub stars. Despite the lack of widespread community testing, it is actively maintained (last pushed 0 days ago) and follows open-source best practices by including a description and utilizing the permissive MIT license.

Verdict
Use with caution — it is safe from a code perspective, but its newness and destructive in-place editing mean you should manually back up your video projects before testing.
SUMMARY

AI Capcut MCP server for automated "talking head" video processing. Removes pauses, detects duplicate takes, adds subtitles — and exports to CapCut project.

README.md

SmartCut MCP Server

MCP server for automated "talking head" video editing. Reads CapCut's auto-generated subtitles to find and remove silences and duplicate takes — directly in the CapCut project.

How it works

  1. Record your video, import into CapCut
  2. Generate subtitles in CapCut (Text → Auto Captions)
  3. Close CapCut
  4. Ask Claude: "Smart cut my 'Podcast Episode 5' project"
  5. Reopen in CapCut, review the cuts

SmartCut reads CapCut's subtitles to understand where speech is. Gaps between subtitles > 1 second are cut. If the speaker repeats a phrase (duplicate take), earlier attempts are removed and the last version is kept.

No API keys required — heuristic analysis works locally. Optionally set OPENAI_API_KEY for GPT-enhanced duplicate detection.

Requirements

  • Python 3.10+
  • CapCut (for editing and subtitle generation)

Installation

cd capcut-ai-editor
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -e .

For optional OpenAI-enhanced duplicate detection:

pip install -e ".[openai]"

Claude Code / Claude Desktop Setup

Add to MCP config:

{
  "mcpServers": {
    "smartcut": {
      "command": "/path/to/capcut-ai-editor/venv/bin/python",
      "args": ["-m", "smartcut.server"]
    }
  }
}

Optional env for GPT-enhanced duplicates:

"env": {
  "OPENAI_API_KEY": "sk-..."
}

Usage

List projects:

Show me my CapCut projects

Inspect a project:

Open CapCut project "My Vlog"

Smart cut (main function):

Smart cut my "Podcast Episode 5" project

With OpenAI enhancement:

Smart cut "My Video" with use_openai=true

Warning: Smart cut modifies the project in place (no backup). Make a copy in CapCut first if you want to keep the original.

Environment Variables

Variable Required Description
OPENAI_API_KEY No OpenAI API key (for GPT-enhanced duplicate detection)
CAPCUT_DRAFTS_DIR No Path to CapCut drafts folder (auto-detected)

Troubleshooting

"No auto-generated subtitles found"
Open the project in CapCut, select the video track, use Text → Auto Captions, save, then try again.

CapCut doesn't see changes
Restart CapCut. It monitors the drafts folder but sometimes needs a restart.

Reviews (0)

No results found