MCP, the Model Context Protocol, is an open standard that lets an AI assistant use another app's tools. Claude, ChatGPT, Grok and many coding assistants support it. For video editing, it means you can give your assistant a real editor to work in, instead of asking it to describe an edit for you to carry out yourself.
What changes with MCP
Without a connection, an assistant can only talk about your video. Connected to an editor over MCP, it can open your project, read what is on the timeline, make the edits and show you the result. You keep the assistant you already know, and the editor does the video work.
Arsaze's MCP server at a glance
- Server URL: https://api.arsaze.com/mcp
- Connection: remote MCP over Streamable HTTP.
- Sign-in: OAuth 2.1. You approve each assistant once with your Arsaze account. There are no API keys.
- Listed in the MCP Registry as io.github.AnrosPrac/arsaze-mcp.
- Works with Claude, ChatGPT and Grok, and with other clients that connect to remote MCP servers with OAuth, such as Gemini, Codex, Cursor, Copilot and Windsurf.
What an assistant can do
- Projects and media: list and create projects, search your library and add media.
- Timeline editing: add, trim, split and ripple-delete clips, add transitions and titles.
- Transcript editing: cut silences, filler words and retakes by editing the text.
- Captions, color and audio: word-timed captions, color wheels, curves and LUTs, loudness, noise cleanup and ducking.
- Looking at footage: check what is on screen at the moments it needs.
- Generation, on paid plans: video, images, voiceover, music and sound effects, placed on the timeline.
- Review and export: share review links, read the comments and make the fixes, and export the video.
Why a timeline and not a render command
An assistant could also run a command-line tool and produce a file. That works for one fixed job. A finished video is hundreds of small decisions that people want to see and change, so Arsaze gives the assistant the same timeline a human editor uses. Every edit plays back live, lands in version history and can be changed by hand.
Room for more than one agent
People and agents can work on the same video at the same time, on separate branches. A branch is a private copy of the timeline, so one agent can try a tighter cut while another works on the captions, without touching the main edit. A person approves a branch before an agent can merge it. You can also lock a clip you are happy with, and no edit, from you or an agent, can change it.
Teach your assistant the editor
Arsaze publishes skills for your AI: short guides on reading a project, cutting by transcript, running a full edit and handling review notes. Upload them to Claude as skills, or add them to the instructions of a ChatGPT project or custom GPT. A connected assistant can also read them straight from Arsaze.
Privacy
An assistant only works on the projects in the Arsaze account you signed in with, and you can remove it from your assistant's settings at any time. In a local project in the desktop app, your files stay on your computer while the assistant edits, and AI access can be switched off for a project or for a whole computer.
Get started
Pick your assistant: connect Claude, connect ChatGPT or connect Grok. The full reference is on the video editor for AI agents page, and a worked example is in what an AI agent does during a real edit.
Questions people ask
Is there an MCP server for video editing?
Yes. Arsaze runs a remote MCP server at https://api.arsaze.com/mcp that lets an AI assistant edit a real multi-track video timeline: cuts, captions, color, audio, generation, review and export.
What does MCP mean in video editing?
MCP, the Model Context Protocol, is an open standard that lets an AI assistant use an app's tools. In video editing it lets an assistant such as Claude or ChatGPT open your project in an editor and make the edits itself, instead of only describing them.
Which AI assistants can connect to Arsaze over MCP?
Claude, ChatGPT and Grok connect as custom connectors, and any client that connects to remote MCP servers over Streamable HTTP with OAuth 2.1 can use the same server URL, including Gemini, Codex, Cursor, Copilot and Windsurf.
Do I need to write code to use MCP video editing?
No. You paste one URL into your assistant's connector settings and sign in to Arsaze. Developers building their own agents can use the same server from any MCP client.
Can several AI agents edit the same video?
Yes. In Arsaze each agent can work on its own branch, a private copy of the timeline, and a person approves a branch before it is merged into the main edit.