YouArt MCP
Connect Cursor, Claude, ChatGPT, Codex, and other compatible AI clients to YouArt.
https://mcp.youart.ai/mcpHow to connect
Paste the URL to your agent
Copy the URL above, drop it into your agent, and ask it to connect. It sets up the MCP server for you.
Sign in and approve
Your agent opens YouArt in a browser to approve the requested permissions.
The exact configuration
Handing over the URL is enough for most agents. If yours keeps asking, or you would rather do it by hand, this is what it should end up with — no API key and no token anywhere.
Add to ~/.cursor/mcp.json
{
"mcpServers": {
"YouArt": {
"url": "https://mcp.youart.ai/mcp"
}
}
}Cursor prompts you to sign in — approve the requested permissions in the browser.
Connecting from SSH or a headless machine?
Authorization always finishes in a browser, and the client waits for the result on a loopback port of the machine it runs on. An IDE remote session (VS Code or Cursor Remote-SSH) handles both for you. A plain SSH session does not — the client prints a URL and then stalls.
- Claude Code: run claude mcp login YouArt --no-browser over ssh -t. Approve in any browser, then copy the URL you land on and paste it back at the prompt. That page shows a connection error, which is expected — the URL itself is what completes the login.
- Codex: it has no paste prompt, so the callback port has to be reachable from the browser. Pin the port by putting mcp_oauth_callback_port = 5555 in ~/.codex/config.toml — otherwise Codex picks a fresh random port every login and there is nothing to forward ahead of time. Then hold a tunnel open from the machine with the browser: ssh -L 5555:localhost:5555 you@your-host.
What YouArt MCP brings to your AI agent
Connect once and your agent can create, build, and run — all inside YouArt.
Generate across image, video, and audio
Create and edit images, animate them into video, write scripts, and produce voiceovers or music — by running YouArt models straight from your chat.
Build and run real workflows
Your agent assembles multi-step workflows on the canvas — adding nodes, wiring them together, and running the whole pipeline, not just one-off prompts.
Reuse your library, or bring your own
Pull any past YouArt generation into a new creation as a reference or start frame, and upload your own images, video, or audio.
One secure connection, no API keys
Sign in once with OAuth and keep the same account, credits, and projects you already use. There are no tokens to copy, rotate, or leak.
Use cases
UGC ads from a single prompt
Describe the product and your agent casts a presenter, generates every shot with its own dialogue, and hands back a finished ad — no filming, no studio, no editor.
Burnt-in subtitles, wiped
Hand over a clip and your agent builds the workflow itself — loading the video, erasing the hardcoded captions frame by frame, then running the whole pipeline.
Just ask — here's what you can do from chat
Talk to your agent in plain language. A few ways to start:
Generate without leaving chat
“Create a moodboard of a desert-glass perfume bottle — 4 variations, studio lighting, 1:1.”
Turn stills into video
“Animate this product shot into a 5-second cinematic clip.”
Build a multi-step workflow
“Write a 3-shot ad script, generate each shot, then stitch them into one video with music.”
Organize your projects
“List my YouArt projects, then add three new concepts to my Q3 campaign.”
Reuse your best work
“Use my last three renders as references for a new shot.”
The latest models, one connection
Your agent reaches YouArt's full catalog. New models become available automatically as they land.
Nano Banana Pro
GPT Image 2
Seedance 2.0- Gemini 3 Pro
Kling 3- Veo 3.1
Flux 2
ElevenLabs v3