Is there an AI agent that can edit videos?

Yes. Valmera’s own agent edits uploaded footage in Studio, and any MCP-capable assistant, such as Claude, ChatGPT or Cursor, can edit through Valmera’s MCP server: upload, cut, caption, reframe, render a preview, check it and export the final MP4.

Can Claude edit videos?

Yes, when it is connected to an editing service. Add Valmera as a custom connector in the Claude app, or run “claude mcp add --transport http valmera https://valmera.io/mcp/server” in Claude Code, sign in, and ask Claude to edit your footage.

Can ChatGPT edit videos?

ChatGPT can edit real footage when it is connected to Valmera as an MCP connector. It then calls Valmera’s editing tools, renders previews and exports the MP4. The ChatGPT setup guide covers the connection steps.

Can my AI agent export the final video?

Yes. It calls export_final with the reviewed project and version, follows wait_for_job, and gets the file from download_url. You can also export the same edit from Studio.

Does my assistant use Valmera’s own agent?

No. Over MCP your assistant is the planner and calls the editing tools directly. Handing a request off to Valmera’s in-house agent is not available over MCP; use Studio for that.

What do I need to connect an AI agent?

A Valmera account with an eligible subscription, which includes MCP access, and an MCP client that supports remote Streamable HTTP servers with OAuth. Clients without OAuth can use a bearer token created at valmera.io/mcp.

Can I edit in Studio while my agent works on the same project?

No. Editing one project from Studio and over MCP at the same time is refused in both directions, so the two can never overwrite each other. Finish in one place, then continue in the other.

AI Agent That Edits Video: Connect Claude, ChatGPT or Cursor

Yes, an AI agent can edit your video. Connect Claude, ChatGPT, Cursor or any MCP client to Valmera’s MCP server, sign in, and your assistant uploads footage, reads the transcript, cuts, captions, reframes, renders a preview, watches it and, when you ask for the finished file, exports the MP4. Your agent does the planning; Valmera supplies the editing tools, rendering and storage.

Prefer no setup? Valmera’s own agent does the same job inside Studio; see how Valmera’s video editing agent works. Both run on the same AI video editor.

MCP endpoint (Streamable HTTP, OAuth 2.1)
https://valmera.io/mcp/server
Claude Code
claude mcp add --transport http valmera https://valmera.io/mcp/server

Two Ways to Have an AI Agent Edit Your Video

	Valmera’s agent in Studio	Your assistant over MCP
Who plans the edit	Valmera’s own agent	Your assistant: Claude, ChatGPT, Cursor, Codex, VS Code or another MCP client
Where you chat	In Studio, next to the preview and timeline	In the assistant you already use
Setup	None beyond an account	Add the connector once and sign in with OAuth
Tools	Valmera’s editing tools, plus the agent’s own planning	The direct editing tools plus session tools: 143 tools (live server card, October 2026)
Export	Export MP4 in Studio	`export_final` → `wait_for_job` → `download_url`
Best for	Hands-off editing in the browser	Folding video into a bigger job: scripts, docs, code or batches

Over MCP your assistant calls the editing tools directly; it does not hand the job to Valmera’s in-house agent. Both routes edit the same kind of project and produce the same full-resolution MP4.

How to Connect Your AI Agent and Edit a Video

1
Add the connector
In the Claude app, add a custom connector with https://valmera.io/mcp/server. In Claude Code, run the command above. ChatGPT, Cursor and other clients have their own setup guides.
2
Sign in
The client opens a browser sign-in (OAuth 2.1 with dynamic client registration and PKCE, scope valmera.edit). Clients without OAuth can use a bearer token from valmera.io/mcp.
3
Create a project and upload
Ask your assistant to upload the file. It calls create_project, upload_start and upload_finish, then index_status until the transcript, shots and silences are ready.
4
Ask for the edit
Describe the result. Your assistant reads the transcript and frames and calls the editing tools; each call writes a new version of the edit.
5
Review the preview
It renders with render_preview and can watch the result with watch_video. Ask it for fixes in the same conversation.
6
Export the MP4
It calls export_final with the reviewed version, follows wait_for_job, and hands you the file from download_url.

Prompts to Give Your Assistant

“Upload interview.mov to Valmera, cut the silences and filler words, add clean captions, and send me the preview.”

“Read the transcript, pick the three strongest complete stories, make each a 9:16 short with captions, and export all three.”

“Record a demo of our website: open the pricing page, scroll to the plans, and zoom in on each click.”

“Watch the preview and list anything that looks or sounds wrong before you export.”

“Export the version I approved as the final MP4 and give me the download link.”

Client-specific walkthroughs: edit video with Claude, edit video with ChatGPT, Claude Code and Cursor. The same plain-English phrasing works here as in editing video by chatting with AI, and the idea behind it is explained in what an agentic video editor is.

What Your Agent Can Do: Workflows and Tool Sequences

Workflow	Typical tool sequence
Clean up a recording	`cut_silences` → `remove_filler_words` → `get_kept_transcript`
Caption it	`add_captions` → `set_caption_style` → `audit_captions`
Make a vertical version	`auto_reframe` → `render_preview` → `look_at`
Podcast to shorts	`make_shorts` → `shorts_status` → `open_short` → `apply_short_edit_batches`
Product demo from a URL	`record_website_demo` → `showcase_demo` → `enhance_cursor`
Add b-roll	`research_broll` → `add_stock_media` → `add_overlay`
Review and deliver	`render_preview` → `watch_video` → `export_final` → `wait_for_job` → `download_url`

For podcast shorts over MCP, your assistant picks the story arcs itself and passes their source ranges to make_shorts, which only creates the child projects, with no captions or framing yet. Your assistant then edits and exports each short; apply_short_edit_batches can save edits for up to 30 shorts in one request. The full clipping workflow is on the AI clipping agent page, and every tool is in the tool reference.

Rules the Server Enforces

The uploaded file is never modified. Every edit is a new version of the edit decision list, and any cut can be restored.
Slow work, such as renders, returns a job id. wait_for_job reports progress instead of a made-up completion.
Every editing call names an explicit project, so an agent never edits the wrong one by accident.
Editing one project from Studio and over MCP at the same time is refused in both directions.
Tools whose backing service isn’t configured are hidden from the tool list instead of failing mid-edit.

See a Real MCP Edit

In the trim-and-reframe example (recorded September 6, 2026), a connected assistant kept three four-second ranges from a 25-second motion graphic and applied a centered square crop through MCP tools. The page shows the original source, the exact edit and the exported file, with an evidence record of versions and file hashes.

Connect Your Assistant to Valmera

MCP access is included with eligible subscriptions. Creating an account and uploading are free.

Create an account →

See pricing →

Service	What the agent can do
Rendley MCP	An assistant builds an editable project, can generate missing media, and returns an exported video
Descript MCP	Imports, Underlord edits, project management and publishing inside a Descript Drive
Reap MCP	Video actions advertised on its MCP page; its documentation server is a separate, docs-only endpoint
Vidrush Rush Agent	An in-editor assistant for targeted changes such as replacing footage or narration

Limits

Over MCP, your assistant can’t delegate to Valmera’s in-house agent; it edits with the tools directly.
It edits footage you provide; it is not a text-to-video generator.
Captions are burned in, with no SRT or VTT export.
No direct publishing to YouTube or TikTok: your agent hands you the MP4.
Quality of the plan depends on the assistant you connect; review the preview before export.

Frequently Asked Questions

Sources

Valmera live MCP server card, checked October 6, 2026
Rendley MCP getting started, checked September 9, 2026
Vidrush Rush Agent documentation, checked September 9, 2026
Descript MCP guide, checked September 8, 2026
Reap MCP, checked September 8, 2026

Made with Valmera: real exports

Let Your Agent Edit the Video

Connect Claude, ChatGPT or Cursor and edit real footage from the conversation. Creating an account and uploading are free; editing requires a subscription.