Skip to content
Agentic Workflowdocs
v0.8.2Install free

use the app / troubleshooting

Performance Optimization

Make your workflows faster and more reliable by reducing delays, heavy operations, and unnecessary steps.

If a workflow feels slow, freezes the page, or sometimes fails, it is usually not a bug.
It is often caused by too much work happening at once inside the browser.

This page explains how to make workflows faster, smoother, and more stable, even on complex websites.


AWFlow runs:

  • Directly inside your browser
  • On the current webpage
  • Using your computer’s memory and CPU

This means:

  • Heavy pages = slower workflows
  • Large extractions = more memory usage
  • Too many steps = longer execution time

Good performance starts with simple workflows.


Before changing your workflow:

  • Close unused browser tabs
  • Restart your browser if it has been open for a long time
  • Disable other heavy extensions temporarily
  • Use Chrome or Edge for best performance

See:


Extracting the entire page text or full HTML can be very slow, especially on long pages.

Typical symptoms:

  • Workflow takes many seconds to start
  • Browser becomes unresponsive
  • AI nodes take a long time to answer

Better approach:

  • Extract only what you need
  • Prefer specific sections instead of the full page
  • Avoid full HTML unless absolutely necessary

See:


Using many extraction nodes one after another increases execution time.

Example of a slow pattern:

  • Extract page text
  • Extract links
  • Extract images
  • Extract tables
  • Extract metadata

Better approach:

  • Extract only what your workflow really uses
  • Avoid duplicate extractions
  • Reuse extracted data when possible

Some websites:

  • Load content progressively
  • Update content while scrolling
  • Use heavy animations or scripts

If your workflow starts too early, it will wait or retry.

Best practice: Always add a Wait for Element step before extracting.

See:


Never extract or click immediately after page load.

flowchart TD
    A[Page Opens] --> B[Wait for Element]
    B --> C[Extract Data]
    C --> D[Process or Analyze]
    class A awf-trigger
    class B awf-flow
    class C awf-io
    class D awf-ai

Waiting for the right element is faster than retrying failed steps.


Scrolling triggers content loading on many websites.

Use scrolling only if:

  • Data appears when scrolling
  • The page uses infinite lists
  • Images or items load lazily

Avoid unnecessary scrolling.

See:


AI nodes slow down when they receive large text.

If AI feels slow:

  • Send only the relevant text
  • Avoid full-page input
  • Summarize or filter before analysis

This dramatically improves speed.

See:


Running many heavy steps together can overload the browser.

Better pattern:

  • One main extraction
  • One transformation or AI step
  • One output or action

Simple workflows are faster and more reliable.


Watch for these warning signs:

  • Execution takes more than 10–15 seconds
  • Browser tab freezes temporarily
  • Workflow sometimes succeeds, sometimes fails
  • Laptop fan turns on loudly
  • Results appear only after a long delay

These usually mean the workflow is doing too much.


A fast workflow usually looks like this:

flowchart TD
    A[Trigger] --> B[Wait for Page Ready]
    B --> C[Extract Only Needed Data]
    C --> D[Optional AI or Processing]
    D --> E[Result or Action]
    class A awf-trigger
    class B awf-flow
    class C awf-io
    class D awf-ai
    class E awf-ok

If your workflow is much more complex, consider simplifying it.


  • Prefer visible text over raw HTML
  • Avoid extracting entire pages
  • Add wait steps instead of retries
  • Limit AI input size
  • Test workflows on real pages
  • Keep workflows focused on one task

If the workflow is still slow:

  • Test it on a simpler page
  • Disable other extensions temporarily
  • Split one large workflow into two smaller ones
  • Compare your design with smaller examples or exported workflows from the community

See:

Ask Aria