Getting started with Agents

Building your first Agent takes a few minutes. Here’s how to get from zero to a running workflow.

Step 1: Choose your starting point

Navigate to Agents in the left navigation sidebar. From there, either use Agent Assistant to build an Agent from a natural language description, or build your Agent manually.

Use Agent Assistant

Type a description of what you want to achieve in the What do you want to build? input field and let Agent Assistant build an Agent for you. If you're new to Agents, this is the fastest way to get started.

Build manually

If you have a specific workflow in mind, click the + New Agent button and start from a blank canvas to build your agent step by step.

Use Agent Templates

Navigate to the Templates tab, browse the template library, select the closest match to your use case, and open it in the editor. From there, run the Agent as-is or customize it however you like.

Step 2: Configure your inputs

Every Agent begins with an Input block. Inputs are the values you supply each time you run the Agent: a keyword, a topic, a URL, and so on.

Click the Input block to open its configuration. Add each input you need, give it a clear label (for example, "target_keyword" or "article_url"), choose a type (text, number, list, date range), and indicate whether it's required. These labels become the variable names you'll reference in later steps.

Step 3: Build your pipeline

Add nodes by clicking the + icon on the canvas or dragging a node from the left sidebar.  or dragging a node from the left sidebar. Each node has a configuration panel that opens on the right: set its inputs, reference variables from earlier steps, and label the output so it's available to subsequent nodes.

Common starting patterns:

• Visibility Score  node → Prompt LLM (generate brief) → Slack Message

• Google Search node → Prompt LLM (analyze results) → Prompt LLM (write article) → WordPress (publish to website)

Add Conditionals to branch your workflow based on variable values. Add Iterations to loop a sequence of steps over every item in a list.

Step 4: Test, publish, and run

New Agents start as drafts. Before publishing, run a test by supplying sample inputs. Each step's output will be visible so you can verify the pipeline behaves as expected.

When the Agent is ready, click Publish to make it live. You can then run it manually, trigger it on a schedule, or supply a spreadsheet of inputs via Sheets to process multiple items at once.