AI Generation

Build workflows from natural language — describe what you want and Orchestrator creates the nodes, edges, and configuration for you.

Orchestrator includes an AI-powered workflow generator that turns natural language descriptions into complete workflow graphs. Instead of manually adding and configuring each node, you can describe your automation and the AI builds it for you.

How It Works

Open the AI Generator

In the workflow editor toolbar, click the AI Generate button (sparkles icon). If this is your first time, you'll see a consent dialog — accept to continue.

Describe your workflow

Enter a natural language description of what you want to automate. Be specific about:

  • What triggers the workflow
  • What steps should happen
  • What integrations to use
  • What the expected output is

Example prompts:

  • "When a webhook receives a customer support ticket, use AI to classify the priority, then send a Slack message to #support-urgent if high priority, otherwise create a Linear ticket"
  • "Generate an image with Skytells TrueFusion Pro from a prompt, then upload it to blob storage and send the URL via email"
  • "Scrape a URL with Firecrawl, summarize the content with AI Gateway, and post the summary to Slack"

Watch it build in real-time

The AI streams operations to the canvas in real-time — you'll see nodes, edges, and configurations appear as they're generated. Operations include:

  • Setting the workflow name and description
  • Adding trigger and action nodes with configuration
  • Connecting nodes with edges
  • Positioning nodes on the canvas

Review and adjust

After generation completes, review the workflow:

  • Check that the right actions were selected
  • Verify configuration fields (prompts, channels, model selections)
  • Attach your integrations to each action node
  • Test with a manual run

What the AI Knows

The AI generator is aware of all available actions across every integration plugin:

  • All Skytells actions (predictions, text generation, image/video creation)
  • AI Gateway text and image generation
  • fal.ai media operations
  • Communication actions (Slack, Resend email)
  • Developer tools (GitHub, Linear)
  • Research tools (Perplexity, Firecrawl)
  • Business tools (Stripe, Clerk, Webflow)
  • System actions (HTTP Request, Database Query, Condition)

It understands how to wire nodes together, set up template variables between steps, and configure condition branches.

Tips for Better Results

  • Be specific about integrations: Say "use Perplexity to search" rather than "search the web"
  • Mention branching explicitly: Say "if the score is above 80, send a Slack message; otherwise create a ticket"
  • Include trigger type: Say "when a webhook receives..." or "manually triggered workflow that..."
  • Describe data flow: Say "pass the generated text to the email body" to get accurate template variable wiring

After Generation

The generated workflow is a starting point. You'll typically need to:

  1. Attach integrations — The AI creates the correct action types but can't assign your credential connections. Select integrations in each action node's config panel.
  2. Review prompts and config — Fine-tune the AI-generated prompts, field values, and model selections.
  3. Test with Manual run — Run the workflow to verify each step works as expected.
  4. Adjust and iterate — Modify nodes, add conditions, or re-run the AI generator to adjust the workflow.

You can run the AI generator on an existing workflow to modify or extend it. The AI will add new operations (nodes, edges) on top of what's already on the canvas.

Next Steps

How is this guide?

On this page