HAR to Load Test Script
Turn a recorded HAR file into a load-test script for k6, Artillery, JMeter, or Locust. Every captured request becomes a step — method, URL, headers, and body — so you can replay a real user flow under load. Filter out static assets, scope to one domain, include or strip headers, and redact auth tokens and cookies so secrets stay out of the script. Everything runs in your browser; the file is never uploaded. Note this generates a script to run yourself — it does not run the load test.
How to use HAR to Load Test
- Record the flow in your browser DevTools Network tab, then “Save all as HAR”.
- Drop the .har file in (or paste the JSON).
- Pick a target: k6, Artillery, JMeter, or Locust.
- Tune the options — skip static assets, choose a domain, include headers, and keep “Redact auth” on.
- Copy or download the script, then run it with the real tool (e.g. k6 run script.js) on your machine or CI.
Frequently asked questions
- Does this actually run the load test?
- No. It generates a script only. Running a load test requires the tool’s own runner — k6, Artillery, JMeter, or Locust — which executes on your machine or CI, not in the browser. The tool gives you a correct starting script to run there.
- Are my auth tokens included in the script?
- Not when “Redact auth & cookies” is on (the default): Authorization, Cookie, API-key and similar headers are replaced with a REDACTED placeholder. The HAR itself is parsed locally and never uploaded, but redaction keeps secrets out of the script you share or commit.
- Why are some requests missing from the script?
- By default “Skip static assets” drops CSS, JS, images, fonts, and media so the test focuses on documents and API calls. WebSocket requests are also excluded because they can’t be replayed as plain HTTP. Turn the option off to include everything, or pick a single domain to narrow the scope.
- Do I need to edit the generated script?
- Usually yes. Real flows need correlation — extracting a CSRF token or session ID from one response and feeding it into the next — plus realistic think-time and your target VU/duration. The generated script is a faithful replay skeleton; add correlation before relying on the results.