Skip to main content

Browser Interaction

Browser automation blocks live under Browser Interaction ▸ Browser Commands in the Toolbox.
This page explains how to start a browser session, supply “smart” parameters, control wait behaviour, and type text or passwords safely.


1 · Activating the browser

  1. Open the Browser tab in the right-hand sidebar.
  2. If a session is already running (for instance, from the agent chat), it appears here.
  3. If no session is running, a cloud-hosted browser instance launches automatically.
  4. Every Browser-category block in your flow (Navigate, Click, Type Text, etc.) targets that live session.

2 · Smart parameters (URLs, selectors & keys)

Many input fields in Browser blocks are smart parameters—they accept either an exact, literal value or a short natural-language phrase that Aidolons resolves at run-time.

Where usedExact input examplesNatural-language examples
Navigate → URLhttps://news.ycombinator.comgoogle
Click → selector / target//*[@id='login'] (XPath)
button.login (CSS)
the login button
Press → keysCtrl+A  Entercontrol a  return  CTRL

Notes on Press keys
• Case and spacing are ignored: control, CTRL, or ctrl all map to the Control key.
• Compound shortcuts (Ctrl+A, control a, ctrl-a) are recognised automatically.
• Common names such as enter / return or esc / escape are interchangeable.

Use whichever style—exact or natural language—is most convenient; the engine will interpret it correctly at execution time.


3 · Wait behaviour for Click / Press

Both Click and Press blocks expose a Wait option under Advanced:

ValueEffect
TruePause until the next navigation completes.
FalseContinue immediately after the action.
Auto (default)AI looks at context (form submits, link clicks, etc.) and chooses whether to wait.

Choose True for guaranteed post-navigation stability or False for single-page apps where no reload occurs.


4 · Typing text and passwords

BlockPrivacy note
Type TextSends the literal keystrokes directly to the remote browser.
Type PasswordSame behaviour—no entered value is sent to the LLM at any point.

Even with this safeguard, it's recommended to use dedicated or low-privilege accounts when automating logins.
That practice keeps runs auditable without exposing personal credentials.