Dynamic web responses sometimes place request values in text that is not clean JSON, XML, or selector-friendly HTML. The JMeter Regular Expression Extractor captures a matched group after a sampler runs, so a later sampler can reuse a token, ID, header value, or short text fragment from the same virtual user flow.
An extractor runs after its parent sampler and stores its result in a variable belonging to the current thread. Attaching it directly to the request that receives the value keeps unrelated responses from overwriting that variable before the next request uses it.
A regular expression depends on the response's spelling, quotes, and field order. Structured HTML and JSON are usually easier to maintain with a CSS Selector Extractor or JSON Extractor; regex is useful for mixed text, headers, or a small predictable fragment. A one-user run exposes a missed match before a larger test repeats the same failure.
The starting plan contains the HTTP request that receives the value, the later request that submits it, and a View Results Tree listener. Both requests run in that order in the same Thread Group. The checkout form shown here returns a hidden field named csrf_token.
Steps to extract values with a JMeter Regular Expression Extractor:
- Check that the validation Thread Group has one user and one loop.
A checkout request can create a real order. The validation plan should target a test environment with disposable data.
- Run the plan from Run → Start to capture the form response.

- Inspect the form request's Response data → Response Body in View Results Tree for the value to capture.
The hidden input has name="csrf_token" followed by its value attribute. The HTML Source Formatted renderer makes the markup easier to read; the extractor will match the original response body.
- Add Regular Expression Extractor beneath the form request through its right-click Add → Post Processors menu.
A child of the form request runs only after that request. An extractor beneath the Thread Group would also run after other samplers in its scope.
- Set Name of created variable to csrf_token.
Main sample only and Body are already selected for this response. A value returned in a header needs Response Headers instead.
- Enter the capture pattern in Regular Expression.
name="csrf_token"\s+value="([^"]+)"
The parentheses capture the value without its quotes. This pattern depends on the two attributes appearing in this order; a different response needs a pattern that matches its actual text. Slash delimiters are not required.
- Set Template to $1$.
$1$ stores the first parenthesized group. $0$ would store the whole match, including the field name and surrounding markup.
- Set Match No. (0 for Random) to 1.
This selects the first occurrence. A value of 0 chooses a random match; a negative value creates numbered variables for all matches.
- Set Default Value to TOKEN_NOT_FOUND.
The visible fallback distinguishes a missed match from an empty value. The Use empty default value option remains unchecked.
- Set the later request's csrf_token parameter value to ${csrf_token}.
Name: csrf_token Value: ${csrf_token}The parameter name belongs to the application; the variable name inside ${...} belongs to JMeter. They happen to be the same here.
- Save the test plan with Ctrl+S.

- Clear the old samples with Run → Clear All.

- Run the updated plan from Run → Start.

- Confirm that the later request's Request → Request Body contains the token from the latest form response.
The form and submitted parameter show the same value. TOKEN_NOT_FOUND means the pattern did not match; a literal ${csrf_token} usually means the extractor did not run or the variable name differs. A Debug Sampler can also expose the variable directly.
Related: How to add a Debug Sampler in JMeter - Check that the later request's Response data reports a successful submission.
The checkout response says Checkout complete after accepting the token. A green sample or HTTP 200 alone may still contain an application error. Temporary debug samplers and View Results Tree are suitable for this small run; large load tests need fewer listeners.
Related: How to disable heavy JMeter listeners before a load test
Related: How to run a JMeter test from the command line