Flow Editor
The Flow editor turns the campaign's Tasks step into a visual canvas: instead of a list of form-based tasks, you drag nodes from a palette and wire them into one journey. Switch between Classic and Flow with the toggle at the top of Step 3 — flows and classic tasks are two views of the same campaign, and existing tasks are materialized as nodes when you open the canvas.

The palette
Traffic sources — where the journey begins:
- Search Engine — search a keyword and click your site (full engine settings included).
- Direct Visit — open a URL directly.
- Google Map — interact with a Maps listing.
- AI Task — let AI drive this part.
Interaction: Click, Type, Hover, Move Cursor, Scroll, Upload File.
Navigation & waiting: Navigate, Search, Visit Other Webpages, Go Back, Referrer Visit, Switch Tab, Close Tab, Wait for Navigation, Wait for Element.
Data & logic:
- Extract → Variable — read a value from the page into a variable.
- Extract All → List — read all matching elements into a list (pair with a For-each Loop).
- Set Variable — compute or assign values with a catalog of expression operations.
- Inject JS — run custom JavaScript, and optionally save its return value to a variable.
- If / Else — branch on multi-clause AND/OR conditions.
- Switch / Router — deterministic N-way branching.
- Loop — repeat / until / while / for-each over a list. The loop body automatically returns to the loop node — no back-edge needed.
- Subflow — run a reusable sub-flow to the end, then continue.
- Parallel — run 2 to 5 branches at once, each in its own tab, then continue.
- Weighted Split — probabilistic branching (e.g. 70% one path, 30% another).
- Delay — wait a number of seconds.
Session & utility: Inject Cookie, Set Cookies, Set Local Storage, Set GEO, Solve Captcha, Screenshot, AI.
Canvas: Note — free-text annotations for documenting your flow.
Editing nodes
Click any node to open its settings panel. A mode node (Search Engine, Direct Visit, …) exposes the same settings as its classic-task counterpart — engine, keywords, CTR, session behavior:

Action nodes configure their target element (Auto / CSS / XPath, attribute matching, index, iframe chains — see the known limitation), plus timing, chance, occurrence limits, and error handling.
Formats & examples
The action nodes accept the same input formats as the classic editor:
- Spintax in any typed text —
{hello|hi|hey} theretypes one variant at random. - Cookie injection — paste DevTools JSON or Netscape cookie files, or point at a file/directory. See the full JSON example.
- Target elements — Auto, a CSS selector (e.g.
button.add-to-cart), or an XPath (e.g.//a[contains(text(),"Next")]), with attribute matching, index selection, and nested-iframe chains (see the known limitation). - Set GEO —
latitude,longitude(see GPS Simulation).
Variables
Values read with Extract → Variable (or assigned with Set Variable) are reusable downstream — Type, Navigate, and condition nodes accept {{variable}} placeholders.
Example flow: Extract → Variable reads a product name into product, then a Navigate node goes to https://google.com/search?q={{product}}, or a Type node types {{product}} into a search box.
Pair Extract All → List with a Loop (for-each) to iterate over every matching element — e.g. read all result links into a list, then loop and click each.
To keep the result of an Inject JS node, fill in Save result as variable. The script's return value is stored there, and you reuse it as {{name}} like any other variable.
Conditions
If / Else and Switch / Router build multi-clause conditions with AND/OR — for example URL contains /checkout AND element .error does not exist → take the success branch. Weighted Split routes probabilistically instead (e.g. 70% path A, 30% path B) for realistic behavior variety.
A clause can check:
- URL, Element (exists or not), or a Variable.
- Element text or Element attribute of the element you select.
- Element count: how many elements match the selector.
- Page text or Page title.
- Last step status: whether the previous step succeeded or failed.
Besides the usual text comparisons, you can use Greater than, Less than, Greater or equal, Less or equal (for numbers, such as an element count), and Matches regex (a regular expression).

Retries and error branches
Every action step has an Error handling section in its settings:
- Retries: extra attempts when the step fails (0 to 5).
- Retry delay (s): the wait between attempts (0 to 60 seconds).
- When it fails: Continue the flow or Stop the visit, used when all attempts fail.
To run other steps on failure instead, connect the step's red error handle to them. When an error path is connected, a failure always follows it, and the When it fails choice is ignored.


Sub-flows
A sub-flow is a reusable mini-flow. Create one with New sub-flow in the bar above the canvas, give it a name, and build its steps. Switch between the Main flow and your sub-flows from the same bar.
To use it, add a Subflow node anywhere and pick the sub-flow. It runs to the end in the same browser, with the same variables, and then the flow continues. A Subflow node has the same error branch and When it fails setting as an action step.
Parallel branches
A Parallel node runs 2 to 5 branches, each in its own new browser tab. The branches run one after another, not at the same time, and each starts with the variables as they were at the node. Use Add branch for more. When all branches finish, their tabs close, the variables they changed are merged, and the flow continues in the original tab from the node's done handle.
- Each branch must end on its own. It can't connect back into the main flow.
- Time cap (seconds) (default 300) limits the whole node: once it has passed, the running branch stops before its next step and later branches don't start.

Tips
- Tidy re-arranges the graph automatically.
- Advanced Functions attached to a classic task appear as step nodes after their mode node when you open the flow — the two editors stay in sync.
- A screenshot node can save into a chosen folder with timestamped filenames.
- Flows are saved with the campaign; reopening the campaign restores the canvas exactly.
Known limitation: iframes
Steps that act on an element inside an iframe aren't supported by the current browser engine yet. This applies to workflow steps and to classic Advanced Functions. Such a step fails with "element not found". Its iframe settings are kept, so it will work once support arrives. Workflow validation warns about these steps; until then, target an element outside the iframe where you can.