SmartCut MCP
SmartCut MCP Server
MCP server that edits talking head videos inside a CapCut project: it uses auto-captions to cut silences and repeated takes.
- 123 ★ on the repository
- Last commit
Extended access
Why this level
- Edits the CapCut project file in place without a backup.
- In GPT mode caption text is sent to OpenAI.
Install
- Works in
Manual install
{"mcpServers":{"smartcut":{"command":"/path/to/capcut-ai-editor/venv/bin/python","args":["-m","smartcut.server"]}}}Install first: a virtual environment and `pip install -e .` from the clone. Replace the Python path with your own.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Edits the CapCut project file in place without a backup.
- In GPT mode caption text is sent to OpenAI.
This is third-party code. Review the repository files before installing.
What it does
SmartCut MCP works with CapCut drafts on your computer. It reads the project's auto-generated captions, uses them to find where speech sounds and cuts pauses longer than one second. If the speaker repeats a phrase, earlier attempts are removed and the last take stays. The agent gets tools to list projects, inspect a project and run the main smart cut operation. Analysis is heuristic and local by default, and no keys are needed. For better repeat detection you can enable GPT through `OPENAI_API_KEY` and the `use_openai` parameter. The workflow is: record, import into CapCut, generate auto-captions, close CapCut, ask the agent, then review the result in the editor.
Who it is for. Podcasters and vloggers who edit single-speaker videos in CapCut and tire of cutting pauses and failed takes by hand.
Good fit when
- An hour-long podcast recording is full of pauses and you want a rough cut in one request.
- The speaker re-took one phrase several times and the project should keep the last take.
- The list of CapCut projects needs a quick look and the right one should open from chat.
- A cut draft is reviewed by eye while the agent does the cutting.
Not a fit when
- The project has no auto-generated captions: the tool works only from them.
- You need multi-track editing with transitions and music.
- The original project must not change without a copy: the server makes no backup.
Example request
Smart cut my "Podcast Episode 5" CapCut project: remove the pauses and keep the last take of repeated phrases.MCP
- Transport
- stdio
- Authentication
- not required
Environment variables
OPENAI_API_KEYsecret- OpenAI key for GPT mode when detecting repeated takes.
CAPCUT_DRAFTS_DIR- Path to the CapCut drafts folder, detected automatically by default.
Limitations
The smart cut operation changes the project in place and makes no backup, so copy the project in CapCut before running it. CapCut must be closed during the cut and restarted afterwards, or it may not see the changes. It works only from auto-generated captions; without them the server reports an error. Python 3.10 or newer and installation from a repository clone are required. The drafts folder is detected automatically and can be set with `CAPCUT_DRAFTS_DIR` if needed. GPT mode passes caption text to OpenAI.
How to disable. Remove the `smartcut` block from the MCP configuration and delete the repository clone with its virtual environment. CapCut projects stay as they were after the last cut.
FAQ
How do I prepare a CapCut project?
Import the video, open the Text menu, then Auto Captions, and generate captions, save the project and close CapCut. Then ask the agent to cut the project by name.
What if the server finds no captions?
Open the project in CapCut, select the video track, generate auto-captions again through the Text menu and Auto Captions, save and repeat the request.
CapCut shows no changes after the cut. What now?
Restart CapCut. It watches the drafts folder but sometimes does not pick up edits right away.
Is an OpenAI key required?
No. The heuristic runs locally without keys. `OPENAI_API_KEY` is needed only for the `use_openai=true` mode.
Which operations does the agent get?
Listing CapCut projects, inspecting a chosen project and the smart cut. The cut accepts a `use_openai` parameter.
Discussion
No questions or reviews yet.
Related
MCP serversMCP server, CLI and desktop app that cut a long video or stream replay into vertical clips with captions
CartCut MCP opens a fresh connection on the next call after an editor update and asks the client to refresh its tool list.
An MCP server for video editing built into the free Concat editor: the agent cuts clips, adds titles and captions, sets keyframes and exports video
An unofficial CLI for CapCut and JianYing projects: subtitles, timing, speed, volume and long-to-shorts, from the terminal or through an agent
ClipForge MCP
ClipForge
ClipForge MCP builds a vertical video from a topic or a product link: stock footage, Edge TTS voiceover and captions, no keys, composed with FFmpeg
VidMoat MCP
Vidmoat MCP
Remote MCP server of a cloud video editor: an agent imports footage, edits the timeline, checks frames and renders MP4
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| SmartCut MCP (this card) | no | not required | 123 | |
| capcut-cli | no | 1,114 | gives manual commands for editing CapCut projects; SmartCut finds pauses and repeats from captions on its own | |
| Video Editing Skill | no | 207 | edits short videos from scratch with captions and B-roll; SmartCut cuts an existing CapCut project |
README in short
The README briefly covers how it works, requirements, MCP setup for Claude Code and Claude Desktop, sample requests and an environment variable table. A troubleshooting section closes it.