CapCut MCP: How to Connect CapCut to Claude and ChatGPT
No — CapCut doesn’t offer an official MCP server, and it doesn’t offer a public API for automated editing either. As of 2026 there is no documented way to ask CapCut’s own infrastructure to build a timeline, cut a pause, or export a finished MP4 on your behalf. CapCut’s “Open Platform” exists, but it’s a plugin surface for tools that run inside the editor plus a narrow text-to-video/template API — not an editing API an AI client could drive.
What people actually mean by “CapCut MCP” is a set of community projects that work around this. They don’t automate CapCut’s UI. Instead they write a CapCut draft file programmatically: you describe the edit to Claude, the MCP server constructs a draft folder, and you drop it into your CapCut drafts directory where it opens as a fully editable project. Here’s what exists, how to set it up, and where the approach breaks down.
New to MCP? MCP servers by category covers the broader landscape.
What an MCP connection to CapCut can do
The community servers are built on top of CapCutAPI / VectCutAPI (sun-guannan, Python, Apache-2.0, ~2k stars), an open-source project that constructs CapCut/JianYing draft files over HTTP. Several MCP wrappers sit on that backend:
- Atx-Guy/capcut-mcp-server (TypeScript) exposes 11 focused tools — create draft, add video/audio/text/image/subtitle/keyframe/effect/sticker, save draft, get duration.
- fancyboi999/capcut-mcp covers draft create/read/modify/save plus transitions, filters, masks, and animations.
- MigueDuque/capcut-mcp-server-extended targets Claude Code with presets and parameterized workflows.
- If you’d rather skip CapCut entirely, pireel (1k+ stars, actively developed) is an open-source editor drivable by any agent over MCP — same conversational shape, its own renderer.
With one connected, you can ask Claude things like:
- Assemble a talking-head clip — drop in this footage, add these SRT subtitles, put a title card on the first three seconds, output 1080×1920.
- Batch social variants — same edit, five aspect ratios, different end cards.
- Templated intros/outros — apply your standard branding pass to every draft before human polish.
The output is always a draft folder (files prefixed dfd_). You copy it into CapCut’s drafts directory and open it in the app. Nothing renders a finished MP4 on its own, and nothing touches your CapCut account or cloud projects.
How to set it up
- Run the backend locally. Clone CapCutAPI/VectCutAPI and start it — Python 3.10+ and FFmpeg required. It listens on localhost by default (
http://localhost:9001). - Register the MCP server in your client config (
claude_desktop_config.jsonfor Claude Desktop, or your Claude Code / ChatGPT connector settings), pointing it at the local backend. - Ask for the edit in chat, then copy the generated
dfd_folder into CapCut’s drafts directory and open it. Polish and export inside CapCut as usual.
If you’d rather skip the MCP layer, the same backend exposes a plain REST API — Claude Code can POST directly to endpoints like add_video, add_subtitle, and save_draft, which suits scripted pipelines better than conversation.
The limits that decide whether this is worth it
- Laptop-bound. The backend, FFmpeg, and your drafts directory live on your machine. Close the lid and nothing runs.
- No triggers. Nothing wakes the AI when new footage lands — a scheduled task can poll, but that’s a poll, not a reaction.
- Draft-file only. Effects that depend on CapCut cloud features, and final rendering/export, stay manual.
- Community maintenance. These wrappers track CapCut’s undocumented draft format, which ByteDance changes between versions. Pin versions and test after every CapCut update.
When you want the video work done, not assembled
An MCP server hands CapCut to an AI inside a conversation you start — useful for drafting, silent until you ask. If the actual bottleneck is that nobody has time to do the editing pass at all, an AI assistant like Carly works differently: connect your email, Drive, or content calendar, and an agent drafts the brief, chases the footage, and coordinates the review loop while you’re away from the keyboard. The MCP route builds tomorrow’s draft; an assistant keeps the pipeline moving around it.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
