Genie and the Node Agent

What stays in the web app

Genie and the Node Agent live only inside the web app: the Genie panel, its ready-made prompts and the Node Agent building a graph on the workflow canvas can't be done through the Scenario MCP. Your own agent can do the same jobs, though: the banner of Fen, the one-change edit and the three-step character chain.

Connect the Scenario MCP

Connect your agent and sign in once at mcp.scenario.com. Then tell it which Scenario team and project to use, so every file lands there.

Give the agent the files

The full-body image of Fen, the reference for every run. Upload it with the agent (it uses upload_asset) or give it its asset ID.

Brief the agent: the banner

The same brief Genie got. The agent writes the full prompt from it, as Genie did, and shows you the plan before it runs.

Use GPT Image 2.5 Sunburst (model_openai-gpt-image-2-5-sunburst), width 3072, height 2048, one image, with referenceImages set to an array holding only the image of Fen.

Brief, word for word:
Store banner for our game: Fen, a fox ranger with a green hood, holding a lantern on a cliff at dusk, painterly, warm light. Landscape 3:2, no text. Use this image for Fen's look. Tell me the model and the cost before you generate.

Write the generation prompt from this brief and end it with "No text, no logos, no lettering". Show me the prompt and the settings. Quote the cost with a dry run first and wait for my go.

Ask for one change

One change per request, and say what must stay. Accept when only the sky changed and Fen, the lantern and the composition are as they were; reject an edit that moved the character or the framing, or let text appear.

Edit the finished banner with GPT Image 2.5 Sunburst (model_openai-gpt-image-2-5-sunburst): referenceImages set to an array with only the banner's asset ID, the same size, and a prompt that asks for a more orange sky and keeps Fen exactly as he is, with everything else unchanged. Quote the cost with a dry run first and wait for my go. Then show me the edit next to the first banner.

The character chain

The Node Agent's pipeline as three runs in a row: generate, remove the background, upscale. Each run takes the previous output's asset ID.

Request, word for word:
Make this a character pipeline for Fen, our fox ranger: generate the character, remove the background, then upscale it.

Generate Fen with GPT Image 2.5 Flare (model_openai-gpt-image-2-5-flare), with the image of Fen as the reference in referenceImages (an array). Then remove the background with Bria Remove Background (model_bria-remove-background), then upscale the cutout with Topaz Image Upscale (model_topaz-image-upscale). Read each model's parameters with model_schema_get first. Quote the whole chain with dry runs first and wait for my go.

File the results and report

Create a collection called "Fen store banner" and add the banner, the edit and the three images of the chain to it. Report the model of each run, every asset ID and the total cost spent.

Open Genie

Click the blue sparkle button in the top bar, or press Cmd+K (Ctrl+K on Windows and Linux). The panel opens on the right of any page; the model picker under the message box sets the language model Genie thinks with, not the image model.

Ask it anything in Scenario

Ask Genie to recommend a model, to take you somewhere ("Take me to the 3D generator." opens Create › 3D) or how something works ("How do I train a model on my own art?").

Which model is best for a painterly store banner with no text?

Start a new chat

Click + before a real job. Or try a ready-made prompt fills the box with a worked example you can adapt.

Brief it, with a reference

Click the paperclip and attach the image of Fen from your library, then lead with the outcome and ask for the plan before anything runs.

Store banner for our game: Fen, a fox ranger with a green hood, holding a lantern on a cliff at dusk, painterly, warm light. Landscape 3:2, no text. Use this image for Fen's look. Tell me the model and the cost before you generate.

Read the plan

Genie fills the Create form: here GPT Image 2.5 Sunburst, its own longer prompt ending "No text, no logos, no lettering", Fen's image under Reference Images and a 3:2 landscape size. Check it against the brief; the cost is on the Generate button.

Say go

Nothing is generated until you say so. Genie then reports the model, size and reference, and tells you when the image is in your gallery.

Go ahead.

Ask for one change

Ask for one change at a time and say what must stay; Genie runs an edit with the finished banner as the reference. Compare both in the viewer.

Same image, but make the sky more orange. Keep Fen exactly as he is. Go ahead.

New task, new chat

Each message rereads the conversation, so long chats cost more. When the task changes, click + and start again.

Build a process with the Node Agent

Workflows › Create, then the sparkle button at the top of the editor's right toolbar (Agent Mode). Read its node-by-node report, then click Run, top right.

Make this a character pipeline for Fen, our fox ranger: generate the character, remove the background, then upscale it.