CapCut MCP Server
Enables AI assistants to fully control CapCut video editing projects, including creating projects, importing media, adding text/subtitles, managing audio, and editing timelines.
README
🎬 CapCut MCP Server
An MCP (Model Context Protocol) server that gives AI assistants full control over CapCut video editing projects. Create projects, import media, add text/subtitles, manage audio, and edit timelines — all through AI-powered tools.
How It Works
AI Assistant ──MCP Protocol──▶ CapCut MCP Server ──Read/Write JSON──▶ CapCut Draft Files
(Python/FastMCP) (draft_content.json)
CapCut stores every project as a folder with a draft_content.json file. This MCP server reads and writes those files directly. You edit via the AI, then open the project in CapCut to preview and export.
Quick Start
1. Install
# Clone your repository
git clone https://github.com/baizo7/capcut_MCP_server-.git
cd capcut_MCP_server-
# Using pip
pip install -e ".[cli]"
# Or using uv (recommended)
uv pip install -e .
2. Configure your MCP client
Gemini IDE / Antigravity
On Windows, open your Gemini MCP configuration file located at:
%USERPROFILE%\.gemini\config\mcp_config.json
Add this to your mcpServers object:
{
"mcpServers": {
"capcut": {
"command": "python",
"args": ["-m", "capcut_mcp.server"],
"env": {
"CAPCUT_DRAFTS_PATH": "C:\\Users\\LENOVO\\AppData\\Local\\CapCut\\User Data\\Projects\\com.lveditor.draft"
}
}
}
}
Note: After saving, restart the IDE for the new server to boot up.
Claude Desktop
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"capcut": {
"command": "python",
"args": ["-m", "capcut_mcp.server"]
}
}
}
Cursor IDE
Add to .cursor/mcp.json in your project:
{
"mcpServers": {
"capcut": {
"command": "python",
"args": ["-m", "capcut_mcp.server"]
}
}
}
3. Start editing!
Once configured, your AI assistant can use commands like:
- "Create a new TikTok project called 'Summer Vibes'"
- "Import all videos from D:\footage\vacation and add them to the timeline"
- "Add subtitles: 'Hello!' at 0-2s and 'Welcome!' at 2-5s"
- "Trim the first clip to start at 3 seconds"
🛑 Important Limitations & Workarounds
1. Windows Store Python Sandbox
If you are running the MCP server using Python installed from the Microsoft Store (which is common in Gemini IDE and Cursor), Windows automatically sandboxes writes to AppData\Local.
- The Issue: The AI will create projects perfectly, but Windows hides them from CapCut.
- The Fix: We built an automatic "Sandbox Escape" into the server using PowerShell! However, for best results, we highly recommend changing CapCut's default save location (Settings -> Project -> Save to) to your
Documentsfolder, and updating yourCAPCUT_DRAFTS_PATHenvironment variable to match it.
2. CapCut Does Not "Hot-Reload"
Because we are editing the raw JSON save files on your hard drive behind CapCut's back, CapCut does not instantly update its screen while you are actively editing a timeline.
- To see changes: You must click the "X" in CapCut to return to the Home screen, then click the project again to force CapCut to reload the timeline from the hard drive.
- For new projects: Simply switch tabs (e.g., from "Space" to "Home") to refresh the project list.
Configuration
| Environment Variable | Description | Default |
|---|---|---|
CAPCUT_DRAFTS_PATH |
Path to your CapCut drafts folder | Auto-discovered |
Auto-discovery
The server automatically checks these locations for CapCut projects:
%LOCALAPPDATA%\CapCut\User Data\Projects\com.lveditor.draft%LOCALAPPDATA%\CapCut\User Data\Drafts
Available Tools (25+)
📁 Project Management
| Tool | Description |
|---|---|
list_projects |
List all CapCut projects on this computer |
get_project_info |
Detailed project info: tracks, materials, segments |
create_project |
Create a new empty project (with canvas presets) |
duplicate_project |
Clone a project as a template |
delete_project |
Delete a project (requires confirmation) |
list_canvas_presets |
Show available canvas sizes (landscape, portrait, etc.) |
🎥 Media Import
| Tool | Description |
|---|---|
add_video |
Import a video file to the timeline |
add_image |
Add an image as a still clip |
add_media_batch |
Batch-import all media from a folder |
list_materials |
List all assets in a project |
remove_material |
Remove an asset and its timeline clips |
📝 Text & Subtitles
| Tool | Description |
|---|---|
add_text |
Add a text overlay at a specific time |
add_subtitles |
Bulk-add subtitles from a structured list |
update_text_style |
Change font, color, size, alignment |
update_text_content |
Change the text content |
🔊 Audio
| Tool | Description |
|---|---|
add_audio |
Add background music or sound effects |
set_volume |
Adjust volume of a specific clip |
mute_track |
Mute all clips on a track |
unmute_track |
Restore volume on a muted track |
✂️ Timeline Editing
| Tool | Description |
|---|---|
trim_clip |
Trim a clip's start/end point |
split_clip |
Split a clip into two parts at a timestamp |
move_clip |
Move a clip to a new timeline position |
delete_clip |
Remove a clip from the timeline |
reorder_clips |
Arrange clips sequentially with optional gaps |
set_clip_speed |
Change playback speed (slow-mo, fast-forward) |
Safety Features
- Automatic backups: Every edit creates a
.backupof the draft before modifying - Non-destructive: Delete operations require explicit confirmation
- Validation: All file paths are verified before import
Development
# Install in development mode
pip install -e ".[dev]"
# Run tests
python -m pytest tests/ -v
# Test with MCP Inspector (browser-based)
npx -y @modelcontextprotocol/inspector python -m capcut_mcp.server
Important Notes
⚠️ Unofficial: This server reverse-engineers CapCut's project format. It is not affiliated with ByteDance.
⚠️ Backups: The server automatically backs up projects before modification, but always keep your own backups of important projects.
⚠️ ffprobe: For accurate video/audio duration detection, install FFmpeg and ensure ffprobe is on your PATH. Without it, the server uses default duration estimates.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.