# Create with VS Code

> Ask the VS Code agent to build a Nodaro workflow, and it appears on your canvas, ready to run again. A real example turns a product description into a video ad.

Source: https://nodaro.ai/docs/mcp/create/vs-code

**VS Code** can build Nodaro workflows for you in agent mode, through the Nodaro MCP server. You describe the steps in plain words. The agent saves them as connected nodes on your Nodaro canvas, and runs them. This page follows one real request: a workflow that turns a product description into a product photo and an 8-second video ad.

To connect VS Code first, follow [Connect VS Code](https://nodaro.ai/docs/mcp/connect/vs-code). It takes one install link and a sign-in with your Nodaro account.

## The request

```
Build me a Nodaro workflow that turns a product description into a product photo and an 8-second video ad, then run it once.
```

We ran this example from Claude Code, through the same Nodaro MCP server that VS Code connects to. The VS Code agent calls the same tools, with the same models, settings and prices.

## The workflow the agent built

The agent called `create_workflow` once, with three nodes and two connections:

Workflow: The workflow on the canvas: the Product text becomes the photo's prompt, and the photo becomes the video's first frame.

- Product → Product photo (prompt)
- Product photo → Video ad (start frame)

- **Product** is a Text node with the description: "Product photo of a brass desk lamp with a pleated linen shade on a walnut desk".
- **Product photo** adds the same style to every description, with [prompt post text](https://nodaro.ai/docs/concepts/prompt-pre-post-text): "soft evening window light, shallow depth of field, clean minimal styling, premium e-commerce photography, no text, no logos".
- **Video ad** opens on the photo and asks for a slow dolly-in, a quiet evening room and soft piano music.

Because the style lives in the node, the description stays short. Change the Product text, and the next run makes a different product in the same style.

## The run

The agent then called `run_workflow`. Nodaro ran the nodes in order, and each node passed its result to the next one.

| Node | Model and settings | Time | Credits |
| --- | --- | --- | --- |
| Product photo | [Nano Banana 2](https://nodaro.ai/docs/models/image/nano-banana-2), 16:9 | 64 seconds | 22 |
| Video ad | [VEO 3.1 Fast](https://nodaro.ai/docs/models/video/veo-3-1-fast), 8 seconds, 16:9, 720p, sound on | 96 seconds | 165 |

Building the workflow cost nothing. Only the two nodes that made a result cost credits.

  
![A brass desk lamp with a pleated linen shade on a walnut desk by a window, in soft evening light](https://nodaro.ai/docs-media/mcp/create/vs-code/lamp-photo.webp)

The product photo: Nano Banana 2, 1376×768

Video: [The video ad: VEO 3.1 Fast, 8 seconds, 1280×720, with sound](https://nodaro.ai/docs-media/mcp/create/vs-code/lamp-ad.mp4)

## Run it again with another product

The workflow is saved, so the next product is one call. Ask the agent, or call `run_workflow` with a new text for the Product node:

```json
{
"workflow_id": "<your workflow id>",
"inputs": {
"product": "Product photo of a walnut cutting board on a marble counter"
}
}
```

The key in `inputs` is the node's id, and the value replaces the node's text for that run. Your own code can run the same workflow through the [Nodaro API](https://nodaro.ai/docs/developers/api/workflows), so a new product page can get its photo and its video ad without anyone opening the editor.

## Open it on the canvas

Every workflow made through MCP is saved in a project named mcp in your Nodaro projects. Open the workflow there, and you see the three nodes, their connections and the results of the run. Change a model, edit a prompt or add a node, and the agent sees your changes the next time it reads the workflow. [Projects and workflows](https://nodaro.ai/docs/mcp/tools/projects-and-workflows) lists the tools that read and change workflows.

## Connect VS Code

1. Open the **Install in VS Code** link on [Connect VS Code](https://nodaro.ai/docs/mcp/connect/vs-code). VS Code opens with Nodaro filled in.
2. Confirm the install, start the server and sign in with your Nodaro account.
3. Switch the chat to agent mode and describe the workflow you want.

The same tools work in [Claude](https://nodaro.ai/docs/mcp/create/claude), [ChatGPT](https://nodaro.ai/docs/mcp/create/chatgpt) and [Cursor](https://nodaro.ai/docs/mcp/create/cursor).

## Frequently asked questions

### Can an AI agent build a Nodaro workflow?

Yes. The create_workflow tool saves a workflow of connected nodes, and the run_workflow tool runs it. The workflow then opens on your canvas like one you built by hand.

### Where does the workflow appear?

In your Nodaro projects, in a project named mcp that Nodaro creates for workflows made through MCP. Open it, and the nodes, the connections and the results are on the canvas.

### How do I run the workflow again with a different product?

Ask the agent, or call run_workflow with an input for the Product node. Your own code can run the same workflow through the Nodaro API.

### How much does one run cost?

Building the workflow is free, and a run costs the credits of its nodes. In the example on this page, one run cost 22 credits for the photo and 165 credits for the video.

### Why build a workflow instead of asking for each result?

A workflow keeps the models, the settings and the prompts that worked. You or your code change one input and run it again, and every run follows the same steps.