1. Docs
  2. Reference
  3. Step reference
  4. Network step

Network step

A page can look right while the call behind it failed. The Network step watches the browser’s own requests during a web test, so you can check the API answer of a click, take a value from it, or list every call and JS error for the report.

6 min read Updated CustomersAdmins

In short

On this page you can

  • Check the call a click made: status code, a response field, a header or the duration.
  • Save a value from a browser call, such as an ID or a token, with Get value.
  • List all calls, failed calls, console messages and JS errors in the report with List.
  • Describe what you want in plain words with Describe with AI.
  • Mark the step as a warning or a failure when calls fail or JS errors appear.

How to use it

  1. Add the step that triggers the call, for example Click Save.
  2. Click Network and choose what it should do: Check, Get value or List.
  3. Under Which call? enter part of the URL, for example /api/orders.
  4. Add rows under Checks or Save to variable.
  5. Save and run; the report shows the matched request and response.

At a glance

Added from
The Network button on the step toolbar. Chips by mode: Net Check, Net Get, Net List.
Project types
WEB only. Not offered in mobile and desktop projects.
Engines
Playwright sees all calls. Selenium sees fetch and XHR calls (and the first page load only in Chrome and Edge).
On failure
A failed check or a missing call: FAILED, the case stops. Failed calls or JS errors: WARNING by default.
Privacy
Authorization, Cookie, Set-Cookie and X-API-Key headers and body fields such as password, token, secret or otp are shown as *** in the log and report. Checks and saved variables use the real values.

The three modes#

ModeAsksWhat it does
CheckDid the call return correctly?Waits for the request made by the previous step and checks the response.
Get valueSave a value from the responseWaits for the request and saves values from it for later steps.
ListReport calls and errorsWrites a snapshot of the important calls so far (API, page loads, errors) with their requests and responses, plus console messages and JS errors, to the report.

Add a Network step#

  1. Put it after the trigger

    Add the step that makes the call first, for example Click Save order. The Network step looks at calls started by the step before it.

  2. Open the dialog

    Click Network. The dialog Add Network step opens. How to use opens a short guide.

  3. What should it do?

    Pick Check, Get value or List. Then choose Set it up myself or Describe with AI.

  4. Which call?

    Enter part of the request URL, for example /api/v1/orders, choose Match and, if needed, Method. For List the URL is optional (“Leave empty to list every call.”).

  5. Checks or variables

    For Check, add rows under Checks (the step fails if one fails). For Get value, add rows under Save to variable. Fill Test Step Name and click Add.

    The step appears with its mode chip. In the right panel it shows a summary with Edit.

Add Network step dialog in Check mode: URL /api/v1/orders, Match contains, a Status code equals 200 check and a Response field check
A Network Check on the orders API.

Which call#

FieldOptionsDefault
URLPart of the request URL. {{variables}} are resolved at run time.— (required for Check and Get value without a prompt: “URL is required”)
Matchcontains, equals, wildcard, regexcontains
MethodAny, GET, POST, PUT, PATCH, DELETEAny

Check#

Each row under Checks has a Target, for some targets a Path / header, an Operator and an Expected value.

TargetPath / headerDefault row
Status code—equals 200
Response fieldJSON path, for example $.data.statusequals …
HeaderHeader namecontent-type contains json
Duration (ms)—less than 2000
Request fieldJSON path in the request bodyequals …
URL—contains …
Call count—at least 1
JS error count—equals 0

Operators: equals, not equals, contains, does not contain, starts with, ends with, greater than, at least, less than, at most, is empty, is not empty, is missing, exists, is true, is false, matches regex. In List mode only Call count and JS error count can be checked.

The step waits for a completed matching call for up to Wait up to (sec) (default 10), because the response may arrive after the click ended. If no call matches in time: “No matching request within 10s”, and the step fails (or is skipped, see Advanced settings).

Get value#

Same call selection as Check; instead of checks you save values. Get value can also run checks.

Save to variable#

Each row has a Variable name (letters, digits and _ only) and a Source. Later steps use {{tc:name}}.

SourceSavesSuggested name
Response field + JSON pathA field of the response, for example $.data.id—
Status codeThe HTTP statusstatusCode
Header + nameA response header—
Duration (ms)How long the call tookdurationMs
URLThe full request URLrequestUrl
Request field + JSON pathA field of the request body (or the whole body)—
JS error count, Console messagesErrors and console output in scopejsErrorCount, consoleLogs
Call count, Call list (List mode)Number of calls, or the list without bodiescallCount, callList
AI result (with a prompt)What the AI foundaiValue

Row errors: “Name is required”, “Letters, digits and _ only”, “Name is used twice”, “Path is required”, and in List mode “List mode supports call count or call list”.

List#

List writes a snapshot to the report: the calls in scope with requests and responses, failed calls, pending calls, console messages and JS errors. By default it uses Request type “API + pages + errors” and Scope “Whole session”. Put a List step at the end of a flow, or after a failure-prone step, to see what happened behind the page. Static files (CSS, scripts, images) are hidden from the list.

Advanced settings#

SettingOptionsDefault
Request typeAPI + pages + errors, Fetch/XHR, Document, AllFetch/XHR (List: API + pages + errors)
ScopeSince previous step: requests started by the step before this one; consecutive Network steps look at the same action. Whole session: everything since the test started.Since previous step (List: Whole session)
If several matchUse the last one, Use the first oneUse the last one
Wait up to (sec)1–120 (“Between 1 and 120 seconds”)10
If no requestFail the step, Skip and continueFail the step
On failed calls / JS errorsMark as warning, Fail the step, Ignore. Covers 4xx/5xx responses, network failures and JS errors in scope.Mark as warning

Describe with AI#

Instead of setting the call and checks yourself, write what you need in Ask in plain words (at most 1,000 characters), for example “The users API returns 200 and there are no JS errors”. At run time the AI picks the calls, values and checks from the calls actually captured; the values themselves are always read from the real data. A URL is optional and narrows what the AI looks at. You can add your own checks under Extra checks. This needs AI enabled for the project. If the AI cannot answer, the report says “AI could not interpret the prompt”.

Limits#

  • Web projects only. A Network step in another project type is refused at run time.
  • Selenium: only fetch and XHR calls; the first page load is visible only in Chrome and Edge. Iframes, service workers, sendBeacon, form posts and static files are not seen. Playwright sees all calls.
  • The network report of a step is fitted into about 60,000 characters; bodies are shortened first, the list of calls stays complete (“Shortened to fit the report”).
  • The captured calls of a test case can be turned into an API test case later, from Maintenance (Network Calls).

If something goes wrong#

Didn’t find your answer?Write to support with your project name and what you see on screen.

info@smartestqa.ai
© 2026 Selenity Technology · SmartestQAsmartestqa.aiThis page as MarkdownChecked against smarttest-fe predev 2a8c79cc, smarttest-be 633c2a03, smarttest-testsuite 017fcec