Zum Inhalt springen
DeutschlandGPT

Getting Started

Build and publish your first workflow step by step

This guide walks you through creating a complete workflow from scratch: a form-triggered pipeline that processes user input through an AI agent and posts a summary to Slack.

Prerequisites

  • Admin or editor access in your workspace
  • A Slack connection configured under Integrations (or substitute any other integration for the example)

1. Create a New Workflow

  1. Open Workflows in the left sidebar
  2. Click New Workflow in the top right
  3. Give it a name, then click Create

The visual canvas opens with an empty trigger slot in the center.

2. Add a Trigger

Click the + placeholder on the canvas, or open the node picker with Space. Select a trigger type:

  • Manual: start by clicking a button; good for testing
  • Form: expose a public form to collect user input
  • Scheduled: run on a cron schedule
  • Webhook: trigger via HTTP POST

For this example, choose Form and add two fields: name (text) and request (textarea). Save the trigger.

You can only have one trigger per workflow. To accept input from multiple sources, build separate workflows and call a shared subworkflow from each.

3. Connect an AI Agent Node

Click the + handle below the trigger node to open the node picker. Under Agent, select AI Agent.

Configure the agent:

  • Model: select your preferred model (e.g. Claude Sonnet)
  • Task: describe what the agent should do. Reference the form outputs:
CODE
Review the following request from {{trigger.outputs.name}} and draft a concise summary with recommended next steps:

{{trigger.outputs.request}}

Leave System Prompt blank for now. Click Save.

4. Connect an Integration Action

Click + below the agent node. Under Integrations → Slack, pick Send Message.

  • Connection: select your Slack connection
  • Channel: choose or type a channel name
  • Message: compose the message using the agent's output:
CODE
New request from {{trigger.outputs.name}}:

{{agent1.outputs.result}}

Save the node.

5. Test the Workflow

Click Test in the top toolbar. The test panel opens on the right.

  • For Form triggers: fill in the test form fields and click Run Test
  • For Webhook triggers: send a sample JSON payload to the shown URL

The canvas shows each node's live status (pending → running → completed). Click any node to inspect its inputs and outputs.

Test runs consume real credits if AI nodes are involved. Use Pinned Data on expensive nodes to replay a cached output during repeated testing.

6. Publish the Workflow

Once the test passes, click Deploy (or Activate) in the top right.

  • Form workflows: a shareable public URL is generated. Share it with users to collect submissions.
  • Webhook workflows: the production webhook URL is shown. Point your external system to this URL.
  • Scheduled workflows: the cron schedule activates immediately.
  • Manual workflows: a Run button appears on the workflow page.

7. Monitor Runs

Open the Runs tab (or the history panel) to see every execution. Each row shows:

  • Trigger time and duration
  • Overall status (completed / failed)
  • Estimated cost
  • Link to the full execution log with per-node details

Next Steps

Was this page helpful?
Getting Started · DeutschlandGPT