Build a simple multi-step workflow in the visual editor, connect an agent to a trigger, and execute your first end-to-end automation in under ten minutes.
Open the Workflow Editor
Click Workflows in the left sidebar, then click "New Workflow". Give the workflow a name and an optional description, then click "Open Editor". The editor is a visual canvas where you drag nodes from the right-side panel onto the canvas and connect them with edges. Nodes fall into three categories: Triggers (what starts the workflow), Actions (what the workflow does), and Conditions (logic branches). The canvas supports pan-and-zoom using your trackpad or mouse scroll wheel, and you can select multiple nodes at once by holding Shift and clicking or drawing a selection rectangle.
Add a trigger node
Every workflow must start with exactly one trigger node. Drag a Trigger node from the panel onto the canvas. Click it to open the configuration panel on the right. Choose the trigger type: Schedule (cron-based), Webhook (receives an HTTP POST), Integration Event (such as a new Slack message), or Manual (run on demand). For your first workflow, select "Manual" so you can run it at will without waiting for an external event. Manual triggers accept an optional JSON input payload that gets passed downstream to action nodes. Leave it empty for now — you can add a payload schema later.
Add an agent action node
Drag an Agent Action node onto the canvas and connect it to the trigger by clicking the trigger's output port (the small circle on the right edge) and dragging to the agent node's input port. Click the agent node to configure it. In the Agent dropdown, select the agent you created in gs-03. In the Input Mapping section, map the trigger's output fields to the agent's input — for example, map {{trigger.payload.message}} to the agent's {{userMessage}} variable. You can also hard-code a test message here. The output of the agent node (the assistant's response text) is available as {{agentNode.output.text}} for downstream nodes.
Run and inspect the workflow
Click "Save" in the top toolbar, then click "Run Now". The canvas enters live-run view: each node lights up green as it completes successfully, or red if it fails. Click a completed node to view its input, output, and execution time in the right panel. If the agent node fails, check the error message for hints — common issues are missing input mappings or hitting a rate limit. After a successful run, click "History" at the top of the canvas to see all past runs with their status, duration, and full logs. You can replay any past run with the same inputs by clicking "Re-run" in the history row.