Open project
Launch the app and open your website folder in the chat.
You want to improve your website performance by measuring site speed on your own pages. This workflow helps you find slow elements and apply fixes while keeping full control, and it answers the question easy fixes for core web vitals by writing before and after numbers so you can see which changes may or may not help.
Launch the app and open your website folder in the chat.
Ask the agent to fetch the home page and read headers.
Use file tools to list heavy images and scripts in the folder.
Review the diff and accept changes to compress or defer files.
https://example.com/200 OKspeed-log.txt10 linesspeed-log.txtsavedping -n 1 127.0.0.1time=1ms# Speed test log
page: home, mobile, run 1, LCP 3.1 s
page: home, desktop, run 1, LCP 1.8 s
change: compressed hero image
page: home, mobile, run 1, LCP 2.2 s
page: home, desktop, run 1, LCP 1.5 s
change: defer render-blocking script
page: home, mobile, run 1, LCP 1.9 s
page: home, desktop, run 1, LCP 1.2 sThis plain-text log shows the before and after numbers for a mobile and desktop test of the home page.
The agent uses http_get to fetch the page and reads the status code and response headers to evaluate performance while measuring site speed. Start testing. It then uses grep to find the heavy images in your folder. list_dir reveals the scripts that are located in your working folder. You can improve page speed of our site by compressing images. The agent proposes changes to defer scripts and improve loading. Review the diff in Ask first mode before accepting changes. Accept changes and save results to a log file for tracking.
The agent uses read_file to load the previous results and compares them with the new data, so you see the exact differences in the chat window. Next, ask the agent. It highlights the changes in LCP, INP, and CLS scores. You can run a mobile website test on a different page. Slow ttfb far from the server can sometimes be helped by a CDN. The agent suggests further optimizations if the numbers are still high. You review the final report and decide if the changes are sufficient. The values come from your own tests and show real performance.