Answers

Which video editors integrate with external LLMs like Claude and ChatGPT? (2026)

Vyra, Descript, DaVinci Resolve Studio 21.1, OpusClip, and Kapwing connect to Claude or ChatGPT over MCP. They differ in whether your AI gets the editing tools or only a prompt box.

By Sulan Zhang, co-founder of Vyra · Updated 2026-10-29

Editors with an MCP server: Vyra, Descript, DaVinci Resolve Studio 21.1, OpusClip, Kapwing (beta)
Clients: Claude Desktop, claude.ai, Claude Code, ChatGPT, Codex, Cursor, depending on the editor
The difference that matters: whether your AI can see the footage and edit the timeline itself
What we checked: each editor's own docs and help pages, October 2026

Real examples from @sulansart

The editors, side by side

Editor	How it connects	Works with	Your AI sees the footage	Your AI edits the timeline	Best for
Vyra	Hosted, headless MCP server, `https://api.usevyra.com/mcp`, OAuth sign-in	Claude Desktop, claude.ai, the Claude mobile app, Claude Code, ChatGPT, Codex, Cursor	Yes, transcript and scene descriptions of every clip	Yes, 50+ editing tools on a real timeline	Turning a folder of raw clips into a finished video
Descript	Hosted MCP server, in Claude's connector directory	Claude, Claude Code through a custom setup	Yes, transcript	Transcribe, trim, remove filler words, add captions, publish	Podcasts and talking videos
DaVinci Resolve Studio 21.1	Built-in MCP server (File, Setup AI Assistants)	Claude, Claude Code, ChatGPT Codex	Through Resolve	Yes, full Resolve	People who already edit in Resolve
OpusClip	Hosted MCP server, `mcp.opus.pro/mcp`, OAuth	Claude, Claude Code	Yes, for picking clips	Exports clips	Cutting one long video into shorts
Kapwing	MCP in beta, by request	Claude	Not stated in its docs	Yes, in Kapwing	Teams already in Kapwing
CapCut	CapCut x Codex	ChatGPT desktop only	Not stated	Not stated	CapCut users on ChatGPT desktop

Open-source FFmpeg MCP servers also exist for developers. They trim, merge, and convert files on your machine, but there's no timeline and the AI can't see what's in the video.

Tools vs a prompt box

There are two ways an editor can plug into your LLM.

Tool-level. The editor exposes its actual operations (add a clip, trim, caption, mask, export) as MCP tools. Your model plans the edit and calls them one by one, and it can check its work by reading the timeline back or capturing frames. Vyra works this way.
Passthrough. Your request goes to the editor's built-in assistant, which does the work its own way. Your model mostly relays the message.

Tool-level matters for three reasons.

You use the model you already pay for and like, Claude or GPT, for the editing decisions.
The model can work across many projects in one conversation, for example captioning 20 clips in the same style.
Developers can script it from Claude Code or Cursor like any other tool.
With a headless server like Vyra's, the edit keeps running on the vendor's servers after you close your laptop.

What Vyra's integration covers

Everything below is available to your model once Vyra is connected. The full list is in the MCP guide.

Area	What your AI can do
See and search	Browse project assets, read the timeline, search footage by what's on screen, what's said, location, or date, capture a frame at any timestamp, search Pexels and Freesound stock
Edit	Add video, audio, and images with trim control, add text, shapes, and styled captions, change position, scale, opacity, volume, and speed, move, split, slip, clone, and delete items, set keyframes
Effects	Apply color and stylize effects, create and animate masks, remove backgrounds, build motion graphics
Generate	Voiceovers, images, sound effects, and music
Project	Create projects, work on several at once, read the project's template and style, add assets from your library, manage markers and tracks

Set it up

Pick your client and follow the steps in the MCP guide. There are separate pages for Claude, ChatGPT, and Cursor, and a developer guide for Claude Code and scripted workflows.

Use Vyra. List the clips in my project "studio vlog" with a one-line description of each. Then propose a 45-second vertical cut with timestamps and wait for my OK before editing.

FAQ

Which video editors work with Claude? Vyra, Descript, DaVinci Resolve Studio 21.1, OpusClip, and Kapwing (beta) all connect to Claude over MCP. Vyra works in Claude Desktop, claude.ai, and Claude Code.

Which video editors work with ChatGPT? Vyra works in ChatGPT on the web, the ChatGPT desktop app, and Codex. Resolve Studio 21.1 works with ChatGPT Codex. CapCut connects through CapCut x Codex on ChatGPT desktop.

Can I use my own LLM subscription instead of paying for the editor's AI? With Vyra, yes. The MCP plans start at $24 a month and use the Claude, ChatGPT, or Cursor subscription you already have.

Does the editor have to be open? Not for Vyra. Its MCP server has been headless since October 2026, so it works with your laptop closed and from the Claude app on your phone.

Do I need to code? No. Hosted servers like Vyra's are added as a connector and you sign in. Claude Code and Cursor are there if you want to script it.

Can the AI see what's in my video? In Vyra, yes. Every clip is transcribed and every scene is described when you upload it, and the AI can capture frames to check its work.

/mcp
/best/mcp-servers-for-video-editing
/answers/video-editing-mcp
/for/developers