Claude + CapCut: What the Integration Can (and Can't) Do in 2026
There’s no official Claude-CapCut connector, and CapCut ships no public API for editing your timeline — but community MCP servers that write CapCut draft files make Claude-driven editing genuinely doable. The way it works isn’t a live plugin inside the app: Claude calls an MCP server that generates a CapCut/JianYing draft on disk, and you open that draft in CapCut to finish or export. There’s also CapCut’s own official “AI API,” but that’s scoped to text-to-video generation and template search, not timeline editing.
Here’s each route, when to pick it, and what to use when you want CapCut work that happens without you at the keyboard.
Route 1: A CapCut draft-file MCP server (the real route)
The most practical way to connect Claude to CapCut is a community MCP server built on the open-source CapCutAPI / VectCutAPI project by sun-guannan (Python, Apache-2.0, ~2k stars, MCP + HTTP). It doesn’t automate CapCut’s UI. Instead it constructs a CapCut draft programmatically: you tell Claude “make a 1080x1920 clip, drop in this video, add these SRT subtitles, put a title on for the first 3 seconds,” and the server writes a draft folder (prefixed dfd_). You copy that folder into your CapCut/JianYing drafts directory and it appears in the app, fully editable.
On top of that backend, Atx-Guy/capcut-mcp-server (TypeScript, ~10 stars) exposes 11 MCP tools — capcut_create_draft, capcut_add_video, capcut_add_audio, capcut_add_text, capcut_add_image, capcut_add_subtitle, capcut_add_keyframe, capcut_add_effect, capcut_add_sticker, capcut_save_draft, and capcut_get_duration. It talks to the VectCutAPI backend running locally (default http://localhost:9001). MigueDuque/capcut-mcp-server-extended is a Claude Code-oriented variant with presets and parametrized workflows, and fancyboi999/capcut-mcp focuses on draft create/read/modify/save plus transitions, filters, masks, and animations.
Setup, roughly:
- Clone and run the VectCutAPI (or CapCutAPI) backend locally — it needs Python 3.10+ and FFmpeg.
- Add the MCP server to your Claude Desktop or Claude Code config (
claude_desktop_config.json), pointing it at the backend. - Ask Claude to build the draft, then copy the generated
dfd_folder into CapCut’s drafts directory and open it.
This is the route to pick if you want Claude to assemble videos from a script — batch social clips, subtitle passes, templated intros — and still do the final polish and export by hand in CapCut.
Route 2: Call CapCutAPI over HTTP directly
If you don’t want an MCP layer, the same open-source project exposes a plain HTTP REST API. Claude Code (or any agent that can run shell/HTTP) can POST to endpoints like add_video, add_audio, add_subtitle, and save_draft. The newer VectCutAPI (also sun-guannan, ~2k stars, v1.5.0) adds cloud-side preview and skills aimed at AI platforms. This is more setup but gives you the full surface without depending on someone else’s MCP wrapper — useful if you’re scripting a repeatable pipeline rather than chatting.
Either way, the output is the same: a draft file you open in CapCut. Nothing here renders a finished MP4 on its own or touches CapCut’s account/cloud.
Route 3: CapCut’s official AI API (narrow, not editing)
CapCut does have an official developer offering — but be clear-eyed about it. As of 2026 there is no public CapCut API for automated rendering, timeline editing, or batch export. CapCut’s “Open Platform” is for building plugins that run inside the editor, and its official “AI API” is scoped to a couple of things: generating video from text and searching/using templates. It’s the ChatGPT-plugin-style surface, thin on docs, and it won’t let Claude edit an existing project. If your task is “generate a short clip from a prompt,” you can point Claude Code at that API with your CapCut developer credentials. For real editing, it’s Route 1 or 2.
Worth noting: CapCut itself is stuffed with AI features (auto-captions, AI voiceover, script-to-video). Those live in the app, not in an API Claude can drive.
The limits that actually matter
- No official Claude/Anthropic connector for CapCut, and no public editing API — everything here is community draft-file generation or CapCut’s narrow AI API.
- No event triggers. Nothing reaches Claude when a file lands in a folder, so it won’t build tomorrow’s clips off the drop. A scheduled task can go look on a timer; that’s a poll, not a trigger.
- Laptop-bound. The backend, FFmpeg, and your CapCut drafts directory all live on your machine. Close the lid and nothing runs.
- Draft, not export. These servers write an editable draft; they don’t render or publish the finished video.
- claude.ai on the web can’t reach any of it — the whole setup is local MCP/HTTP.
Claude is excellent for assembling a CapCut draft from a script on demand. It’s not built for CapCut work that has to happen on a schedule, unattended.
If you want CapCut work that runs on its own: Carly
Claude is a phenomenal in-the-moment editing partner. But if the job is “the same video pipeline, every time something happens, without me babysitting a terminal,” that’s a different tool.
Carly reaches CapCut through your own CapCut API key (paste it on the integrations page) — one of 200+ native integrations plus anything with an API — and runs the surrounding pipeline on real events, 24/7. A few concrete scenarios:
- When a new raw clip lands in Google Drive or Dropbox, Carly kicks off the CapCut generation step, then drops the draft link in Slack for review.
- When a script is approved in Notion, Carly assembles the templated intro/outro and pings the editor that it’s ready.
- When a video finishes rendering, Carly posts it to YouTube or a client folder and emails the client that it’s live.
- No-code setup. You don’t write config. Carly interviews you about the workflow in plain English and wires the steps.
- Actually sends — Carly reaches both Gmail and Outlook, so notifications and hand-offs land in the right inbox.
- Connects to anything — 200+ native integrations, plus any tool with an API via your own key.
AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.
Claude vs Carly
| Claude + CapCut | Carly | |
|---|---|---|
| Build a CapCut draft from a script | Yes (via MCP/HTTP) | Orchestrates the step, not the timeline |
| Act inside a live session you start | Yes | Yes |
| Acts on triggers / schedules | No | Yes |
| Reacts the moment a file lands | No | Yes |
| Sends email (Gmail + Outlook) | No | Yes |
| Reaches 200+ other tools | Via separate MCP servers | Native + BYO-key |
| Official CapCut connector | No | Via your CapCut API key |
| Pricing | Pro $20 / Max $100–$200 | AI agents from $35/mo |
Claude is the best copilot for building a CapCut draft on demand; Carly is the way to make that video pipeline run itself.
Frequently Asked Questions
Does Claude work with CapCut?
Yes, indirectly. There’s no official connector, but community MCP servers built on the open-source CapCutAPI/VectCutAPI project let Claude generate CapCut draft files you open in the app. Claude doesn’t drive CapCut’s UI or export finished video.
Is there an official CapCut MCP server?
No. CapCut doesn’t publish its own MCP server. The real options are community projects — sun-guannan’s CapCutAPI/VectCutAPI (with a built-in MCP mode), Atx-Guy/capcut-mcp-server, MigueDuque’s Claude Code-focused extended fork, and fancyboi999/capcut-mcp — all of which write CapCut/JianYing draft files.
Can claude.ai on the web access CapCut?
No. Every working route relies on a local MCP server or HTTP backend plus your local CapCut drafts folder and FFmpeg. The claude.ai web app can’t reach any of it.
Can Claude edit and export a CapCut video automatically?
Not on its own. These servers produce an editable draft, not a rendered file, and nothing about a new file reaches them. For unattended, trigger-based CapCut pipelines you need an automation layer like Carly.
More: Claude connectors explained · Claude vs Carly · Can Claude send emails? · Claude + Notion · Claude + Slack · All integrations
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."
