← Home
SETUP GUIDE

Published · Updated

How to Connect Claude to Video Editing

Claude cannot open your footage or render a file by itself. Connected to a video editing MCP server it can — the editing engine does the work, and Claude chooses and sequences the tool calls. Setup is one paste and one sign-in.

This guide covers both the Claude app and Claude Code, what Claude can do once connected, and the handful of errors worth recognising.

PASTE THIS INTO CLAUDE
https://entrepreneur-bot-backend.onrender.com/mcp

Connect Valmera to the Claude App

  1. 1
    Create a Valmera account
    Sign up free at valmera.io with email or Google. Every account gets 50 credits, no card required. If you sign up with Google, set a password with “Forgot password” — the connector's consent screen needs one.
  2. 2
    Add the custom connector
    In Claude, open Settings → Connectors → Add custom connector and paste https://entrepreneur-bot-backend.onrender.com/mcp. There is nothing else to fill in — no client id, no secret, no token.
  3. 3
    Sign in and press Allow
    Claude discovers the authorization server from the endpoint, registers itself, and opens a Valmera sign-in page. Sign in with your Valmera email and password, review what you are granting, and press Allow.
  4. 4
    Confirm it worked
    Say “list my Valmera projects”. If you get a list back (or an empty one on a new account), the connection is live and you can start editing by asking.

Setup is one paste and one sign-in. The first upload then needs a one-time analysis pass, which reports progress and is reused by every edit afterward.

Connecting Claude Code

Claude Code can run the same OAuth flow, or carry a bearer token instead. To use a token, mint one from your Valmera session and then add the server:

claude mcp add --transport http valmera \
  https://entrepreneur-bot-backend.onrender.com/mcp \
  --header "Authorization: Bearer vlm_mcp_..."

The token is displayed once and stored only as a sha256 hash — if you lose it, mint another rather than hunting for it.

What to Say Once It Is Connected

“Upload this and cut all the dead air and the ums.”

Claude uploads the file, waits for indexing, then calls silence removal and filler-word removal. Both snap to word boundaries from the real transcript, so nothing sounds clipped.

“Add karaoke captions with a yellow highlight and put chill music under my voice.”

Captions are timed word by word from the transcript; the music is mixed with automatic ducking so it drops under speech and comes back in the gaps.

“Reframe it for Reels and punch in where I get loud.”

Subject-aware reframing samples frames and asks a vision model where the subject is before cropping to 9:16. Punch-ins land on words whose vocal stress was measured from the audio.

“There are old burned-in subtitles at the bottom — get rid of them.”

Claude measures the rectangles from the actual frames, then repaints those pixels and reconstructs the background. The subtitles are gone rather than blurred.

“Render a preview and tell me if the captions collide with anything.”

A preview renders and Claude inspects real frames from it. This is the loop that makes the editing agentic rather than fire-and-forget.

“Export it and give me the download link.”

The final render comes from your ORIGINAL upload at source quality, not the preview proxy. Claude waits on the job and returns a download URL.

If Something Goes Wrong

The sign-in page rejects my password
If your Valmera account was created with Google sign-in, it has no password yet and the consent screen needs one. Set it with “Forgot password” on valmera.io, then reconnect.
It says my account does not have access
That message means exactly what it says — the password was right and the account is not permitted. It is deliberately worded differently from a wrong password, because telling you “wrong password” when the password was correct sends you round in circles.
A tool call says STILL RUNNING with a job number
That is correct behaviour, not an error. Renders, exports, pixel repainting and website recording take longer than a request should block for, so they hand back a job id. Ask Claude to wait for it, or just say “check on that job”.
It refuses because an edit is already running
The same project cannot be edited from the studio and from Claude at once — it is refused in both directions rather than allowed to corrupt the edit decision list. Finish or cancel the turn in the studio, then retry.
A tool I expected is not in the list
Valmera hides a tool whose backing service is unconfigured rather than exposing one that fails when called, so the live tool count can be lower than the published registry. Stock b-roll search and AI image generation are the usual examples.
My video is too large to upload through the connector
MCP arguments are JSON, so raw bytes never cross the protocol. Past roughly 64MB, use the presigned multi-part helper — or simply give Claude a link and let Valmera fetch the media itself.

Frequently Asked Questions

Use an MCP server that exposes a real video editor. In the Claude app, open Settings → Connectors → Add custom connector and paste https://entrepreneur-bot-backend.onrender.com/mcp. Claude discovers the authorization server, registers itself, and opens a Valmera sign-in and consent page; sign in and press Allow. There is no token to copy. From then on you edit by asking — upload footage, cut, caption, mix, render, export, download.
Not on its own — Claude cannot open your footage or render a file. Connected to a video editing MCP server it can, because the editing engine does the work while Claude chooses and sequences the tool calls. With Valmera connected, Claude has 97 editing tools plus 11 session tools and produces a real MP4 you download, not a description of one.
Yes, and Claude Code has a second option. It can run the same OAuth flow, or carry a static bearer token instead: `claude mcp add --transport http valmera https://entrepreneur-bot-backend.onrender.com/mcp --header "Authorization: Bearer vlm_mcp_..."`. The token is shown once and stored only as a hash. Both paths reach the identical tool registry.
Your original upload is never modified. Every tool mutates an edit decision list and a renderer produces video from it, so nothing the agent does is destructive: every operation creates a new version, anything cut can be restored, and final exports are rendered from the original file at source quality. That is what makes it reasonable to let an autonomous agent work on footage you cannot re-shoot.
Nothing — that is the point. The tools are not re-declared for MCP. Valmera's editing engine publishes one registry and the connector serves it verbatim, so the model in your Claude session sees the same names, the same schemas, the same descriptions and the same refusals. There is no second list to drift out of date.
You need a Valmera account. Every account gets 50 free credits on signup with no card required, and paid plans (Creator $30/mo, Pro $50/mo, Frontier $100/mo) open with a 3-day free trial.
It is a standard streamable-HTTP MCP server with OAuth 2.1, so any compliant client should work. The Claude app and Claude Code are the two we have verified end to end — we would rather name those two honestly than list logos we have not tested.

Connect It in Two Minutes

Create a free Valmera account, paste the endpoint into Claude, sign in — then edit your first video from the chat.

Create a free account →
See pricing →

Related Articles

Valmera MCP Server
The server itself: transport, auth, design notes.
MCP Tool Reference
Every tool Claude gets once connected.
Claude as a Video Editor
What Claude can and cannot do with video, answered honestly.
Agentic Video Editor
Why an agentic editor is the kind that can be driven by another agent.