Skip to main content

Generate Response Workflow Action (OpenRouter)

The Generate Response workflow action sends a prompt to a selected OpenRouter AI model and returns an AI-generated response that can be used in later workflow steps.

What the Generate Response Action Does

The Generate Response action uses OpenRouter to access AI models inside a Kyrios workflow.

When the workflow reaches this action:

  1. Kyrios sends the System Prompt, Prompt, and selected model to OpenRouter.

  2. The selected AI model generates a response.

  3. Kyrios stores the generated response as an output variable.

  4. Later workflow steps can use the generated response.

Key Benefits

Generate Response helps workflows create AI-powered text without manual writing.

Benefits include:

  • Access to multiple AI models through OpenRouter

  • Dynamic responses using workflow variables

  • Custom tone and behavior through System Prompt

  • Flexible use in SMS, email, notes, and CRM updates

  • Scalable content generation inside workflows

Connect OpenRouter Before Using Generate Response

OpenRouter must be connected before the Generate Response action can run successfully.

To connect OpenRouter:

  1. Go to Settings.

  2. Select Integrations.

  3. Locate OpenRouter.

  4. Click Connect.

  1. Enter your OpenRouter API key.

  2. Click Save.

Create or Open a Workflow

To add Generate Response, first create a workflow or open an existing workflow.

  1. Go to Automation.

  2. Open Workflows.

  3. Click Create Workflow.

  4. Choose how you want to create the workflow, such as Start from Scratch.

Add a New Workflow Action

After the workflow opens, add a new action step.

  1. In the workflow builder, click + Add.

  2. Select Add Action.

  3. The Actions panel opens on the right side.

Select the OpenRouter App

Generate Response is located under the OpenRouter app.

  1. In the Actions panel, click the Apps tab.

  2. Scroll to OpenRouter.

  3. Click OpenRouter to expand its available actions.

Choose Generate Response

After expanding OpenRouter, select Generate Response. Generate Response runs your prompt through the selected AI model and returns the generated reply to the workflow.

Configure the Generate Response Action

After selecting Generate Response, the configuration panel opens.

The configuration includes:

  • Action Name

  • System Prompt

  • Prompt

  • Model

Use these fields to control what the AI generates and how it should behave.

Action Name

The Action Name identifies this workflow step. Use a clear name that describes what the action generates.

Example:

Generate Lead Follow-Up Email

System Prompt

The System Prompt defines the AI’s role, tone, behavior, and rules. Use this field to tell the AI how it should behave across the request.

Example:

You are a helpful front office assistant that provides clear, professional responses to customer inquiries.

Prompt

The Prompt defines the specific task the AI should complete. Prompts can include workflow variables, contact fields, custom fields, and outputs from earlier workflow steps.

Example:

Write a short follow-up email for {{contact.first_name}} based on their recent form submission.

Model Selection

The Model field controls which OpenRouter model generates the response.

Choose a model based on the workflow need:

  • Faster models for simple responses

  • Stronger reasoning models for complex instructions

  • Creative models for marketing copy

  • Lower-cost models for high-volume workflows

System Prompt vs. Prompt

The System Prompt controls how the AI behaves. The Prompt controls what task the AI completes.

Example:

System Prompt:
You are a friendly sales assistant that writes short and persuasive messages.

Prompt:
Create a follow-up message for {{contact.first_name}} who downloaded our pricing guide.

Use both fields together when the workflow needs consistent AI behavior and dynamic task instructions.

Test the Action

Test the action before saving or publishing the workflow.

To test the action:

  1. Complete the System Prompt, Prompt, and Model fields.

  2. Click Test Action.

  3. Review the generated response.

  4. Adjust the prompts if the result is not correct.

Testing helps confirm the selected model returns the expected output.

Save the Action

After configuration is complete, click Save Action. The Generate Response action is then added to the workflow.

The generated response can be used in later workflow steps, including:

  • Send SMS

  • Send Email

  • Internal notifications

  • Notes

  • CRM field updates

  • Conditional logic

Common Issues and Fixes

OpenRouter does not appear in workflow apps

Connect OpenRouter in Settings > Integrations before adding the workflow action.

The action does not generate a response

Confirm the OpenRouter API key is saved correctly and the selected model is available.

The response is too generic

Add more context to the Prompt and clearer behavior rules to the System Prompt.

The response does not match the desired tone

Update the System Prompt with tone, formatting, and style instructions.

Workflow variables do not populate correctly

Confirm the variable exists earlier in the workflow and uses the correct merge field format.

FAQ

What is OpenRouter?

OpenRouter is the integration used by Kyrios to access multiple AI models through one connection.

Can I use workflow variables inside prompts?

Yes. Prompts can include workflow variables, contact fields, custom fields, and previous action outputs.

Where is the AI-generated response stored?

The generated response is stored as a workflow output variable that can be used in later workflow actions.

Can I use the generated response in SMS or email messages?

Yes. The generated response can be inserted into Send SMS, Send Email, and other workflow actions.

Can Generate Response be used with conditional workflow logic?

Yes. Later workflow steps can use the generated response in conditional logic.

What happens if the model fails to generate a response?

The workflow behavior depends on the configured automation path. Test the workflow before publishing to confirm the expected behavior.

Did this answer your question?