dGEN Visual StudioSign in
dGEN Visual Studio MCP for any AI
Create images and videos directly from your prompts in any AI tool
Already on dGENIX? Connect in one click and automate your studio workdGENIXCopy the connector URL
You paste this into Claude in the next step.
https://dgenvisual.com/mcpGo to Claude, then Connectors
In the Claude desktop app or on claude.ai, open Settings and then Connectors. Add a custom connector, name it dGEN Visual Studio and paste the URL.
Open ClaudeConnect, sign in and start
Claude asks which permissions it wants. You see that on a screen of ours before anything happens, and you approve it there.
Start creatingAdd the server
One command in your terminal, from any folder.
claude mcp add --transport http dgen-studio https://dgenvisual.com/mcpApprove it in your browser
The first time you use a tool, your browser opens our consent screen. After that Claude Code remembers the connection.
Ask for an image or a video
Everything you can do on the canvas, you can now ask for in plain language. The result lands in your library, at the same price.
Start creatingCreate a key
An agent that runs without a browser cannot walk through a consent screen. Create a key under Connected apps. You see the secret once, because we keep only an encrypted fingerprint of it.
Exchange it for a token
Client credentials, the way any OAuth client does it.
curl -X POST https://dgenvisual.com/api/oauth/token -d grant_type=client_credentials -d client_id=$STUDIO_CLIENT_ID -d client_secret=$STUDIO_CLIENT_SECRET -d resource=https://dgenvisual.com/mcpCall the server
Send the token as a bearer on https://dgenvisual.com/mcp. Your agent gets exactly the permissions you granted the key, and nothing beyond them.
Go to Connected appsGive it a job to do
A skill file for any of the three ways above: how this studio works, which model to reach for, how to prompt it, what it costs, and where it stops. Drop it in your assistant’s skills folder, or paste it into its instructions, and it knows the platform without being told twice.
One thing it is good at: giving every article your assistant writes its own cover. That guide is here.
You decide what an agent may do
Permissions are granted separately. An agent that may only look cannot make anything, and if it tries anyway that attempt shows up in your overview.
View your flows and templates
The assistant can see which flows you have and how they are put together. It cannot change or delete them.
Run a flow from start to finish
The assistant can run an existing flow from start to finish. That costs credits, exactly as much as when you press Run yourself. It cannot edit flows.
View and download your files
The assistant can look up your files and get a temporary download link. It cannot delete or overwrite anything.
View your own LoRAs
The assistant can see which LoRAs you have, so it can use one when generating. It cannot add or remove any.
Create images (costs credits)
The assistant can create images. That costs credits per image, the same amount as on the canvas. If one fails through an outage you get those credits back.
Create videos (costs credits)
The assistant can create videos. That is the most expensive action in the studio, so keep an eye on the daily ceiling under Connected apps.
View your credit balance
The assistant can read your balance, so it knows whether something still fits. It cannot buy or change anything with it.
And you can take it back
You see everything
Every action by every connected app is in your overview, refused ones included, with what it cost.
You can stop it
One click on Revoke and the access is gone, including access that was still valid. The agent cannot request new access.
There is a limit
A connection may do at most 30 actions per minute and spend 5,000 credits per day. A runaway agent does not cost you your whole wallet.
The same price as on the canvas
An agent uses exactly the same execution path you do. It cannot get anything cheaper, or outside the rules.