Skip to content
Agentic Workflowdocs
v0.8.2Install free

use the app / workflows / components

Nodes

Understand and use nodes — the building blocks of browser workflows in AWFlow.

Nodes are the steps in your workflow.

Each node does one thing, like extracting text from a page, clicking a button, or formatting data.

Types of Nodes

Section titled “Types of Nodes”
graph TB
  WorkflowNodes[Nodes] --> Browser[Browser Nodes]
  WorkflowNodes --> Processing[Processing / Built‑In Nodes]
  WorkflowNodes --> AI[AI Nodes]
  WorkflowNodes --> Logic[Flow & Logic Nodes]
  class Browser awf-io
  class Processing awf-data
  class AI awf-ai
  class Logic awf-flow
  • Browser nodes — work with the page you’re on (extract text/HTML, click, scroll, fill forms)
  • Built‑in nodes — transform data, call APIs, work with files, etc.
  • AI nodes — summarize, classify, or analyze content
  • Logic nodes — choose different paths (If), wait, handle errors

You can browse all available node types in the Node Reference or by category:

Node runtime markers

Section titled “Node runtime markers”

Some nodes show a small icon in the node list and on the canvas that tells you where the node runs and what it needs:

  • ☁️ Cloud — the node sends your data to the AWFlow backend to do its work (for example the Google integration nodes). It needs a connected account and an internet connection.
  • 🌐 External API — the node calls a third-party service’s API directly from your browser (for example Airtable, Notion, Slack). It sends data to that service and needs an account and internet connection.
  • 🧩 Extension — the node runs inside the AWFlow browser extension (the in-page action nodes). It needs the extension installed.
  • No marker — the node runs entirely in your browser (data transforms, code, and similar). No account or network required.

Use these markers to see at a glance which steps in a workflow leave your device or require the extension.

Adding a Node

Section titled “Adding a Node”

In a new (empty) workflow

Section titled “In a new (empty) workflow”
  1. Click Add first step.
  2. Use the Node Library panel to search or browse nodes.
  3. Click a node (e.g. Get Selected Text, Get All Links).
  4. The node appears on the canvas, ready for configuration.

In an existing workflow

Section titled “In an existing workflow”
  1. Hover between two nodes or at a free connector point.
  2. Click Add node.
  3. Pick a node from the panel.
  4. Connect it to previous and next nodes.

Common workflows often chain patterns like:

  • Extract → Analyze → Output
  • Get Links → Filter → Visit & Process
  • Click / Scroll → Extract → Condition / Branch

Node Controls & Actions

Section titled “Node Controls & Actions”

Hover or right-click a node to see common actions:

  • Execute Step — run that node immediately
  • Deactivate — temporarily disable (without deleting)
  • Delete — remove the node
  • Open / Configure — adjust parameters
  • Rename, Duplicate, Copy, Clear, Pin — help manage and organize your workflow
  • Node Notes — add documentation visible on the canvas

Behavior of Browser Nodes

Section titled “Behavior of Browser Nodes”

Browser nodes behave a little differently because they depend on the page:

  • They operate on the current page state, extracting real-time content
  • They respect browser security rules, cross-origin limits, and sandboxing
  • They are optimized for browser memory and performance
  • They can be configured (timeouts, retries, “always output”, etc.)
  • They may fail if page structure changes or if content is not present

Node Settings & Options

Section titled “Node Settings & Options”

Under a node’s Settings tab, you’ll find options that control its behavior and reliability:

  • Always Output Data — ensure the node emits an (empty) result even if nothing is found
  • Execute Once — run only once per workflow run
  • Retry On Fail — retry browser operations when they fail
  • On Error:
    • Stop Workflow
    • Continue
    • Continue (passing error info downstream)

Browser nodes may have extra settings:

  • Page wait / load time
  • Element timeouts
  • Retry delays
  • Notes and descriptions — helpful for documenting what each node does

Data Flow & Node Outputs

Section titled “Data Flow & Node Outputs”

Nodes pass data to the next node.

Most of the time, the data is shown as JSON in the output panel.

  • A node can emit zero, one, or many items
  • Fields in the items can be manipulated via expressions (e.g. {{ $json.text.length }})
  • Use nodes like Edit Fields to create or reshape output fields

If you want more insight into how data moves between nodes, see Data Structure and Item Linking.

Actual and expected output

Section titled “Actual and expected output”

Every node declares the shape of the data it outputs. The Output tab of a node’s settings shows two views:

  • Expected — what this node will output with its current configuration, before you run it. Use it to write expressions for the next node without a test run.
    • Example shows sample JSON with placeholder values (not real data).
    • Structure lists every field with its type, and marks it as required or optional. Expand objects and lists to see what they contain. “Additional fields” means the node may output more fields than those listed; a note tells you when the node also passes on the fields of the item it received.
  • Actual — the data the node really output in its last run. It opens automatically after a run.

Some nodes cannot describe their output in advance:

  • Finish configuring this node — the output depends on a choice you haven’t made yet, such as the operation of an integration. Choose it in Parameters and the preview appears.
  • No valid content — the output is whatever an external service or your own code returns (for example Custom API or Code), so there is nothing fixed to preview. Run the node to see it.
  • This node does not emit workflow output — the node gives a capability to another node (for example a chat model connected to an agent) or always stops the run.

When a node outputs data that doesn’t match its declared shape, the step fails with Output validation failed. The Actual view lists the field at fault, the type that was expected and the value that was received; Compare with expected structure opens the Expected view next to it. The failure behaves like any other step error: Retry, Continue on fail and Try / Catch all apply.

When a step fails

Section titled “When a step fails”

A step that fails, with no Try guarding it, stops its own branch: the steps after it don’t run. Branches that don’t depend on it keep running, so a node wired in parallel next to a failing step (for example a Display Markdown next to a Copy to clipboard) still runs.

  • A step that joins several branches needs all of them, so it doesn’t run if one of them failed. A Funnel still runs as soon as any branch that didn’t fail reaches it.
  • The run is still reported as failed, with the first step that failed as its cause, and any error workflow runs once.

What’s next?

Section titled “What’s next?”

Using Browser Nodes

Explore all nodes that let you interact with web pages (click, extract, scroll, etc.).

Go →

Built-In / Integration Nodes

Discover nodes for data transformations, HTTP calls, file operations, and more.

Go →

Flow & Logic Nodes

Understand how to build branching, loops, and conditional logic in workflows.

Go →

Ask Aria