Cursor IDE — built-in browser
Version: Cursor 3.23.23 (Chrome 148.0.7778.280, Electron 42.10.0), macOS (Darwin 27.0.0). Updated this morning. A colleague on the previous Cursor build can still open the same app in the Cursor browser.
Steps to Reproduce
Run a local SPA at http://127.0.0.1:5005 that signs in with credentials: include against https://rewards-admin-api-dev.spurcorpgifting.com.
Open that URL in the Cursor browser and sign in.
Expected: The session cookie from the API is stored and sent on the next request, as it is in Firefox on the same machine and in the Cursor browser on the previous release.
Expected Behavior
Actual: Login (POST /brochure) returns 200. The next call (POST /user_role_allowed) returns 401 five times. The app then shows “Your session has ended”. The page is a top-level window, not an iframe. chrome://settings/cookies does not open in this browser, so third-party cookies cannot be allowed from the UI.
Hey, thanks for the detailed report. We checked and found no changes in 3.23.23 to how Cursor’s browser stores or sends cookies. It also uses the same Chromium version as the previous release: Chromium 148.0.7778.280 and Electron 42.10.0.
One thing that often explains “it works in Firefox but not in Cursor” is that Cursor’s browser is based on Chromium. Chromium treats cookies without a SameSite attribute as SameSite=Lax. Lax cookies aren’t sent with cross-site fetch requests, so a page on 127.0.0.1 calling an API on another domain won’t receive its session cookie back. Firefox doesn’t apply that default. Google Chrome on your machine would behave the same way as Cursor.
Could you check this and share what you see?
Open the app in Cursor’s browser.
Right-click the page and select Inspect Element.
Open the Network tab in DevTools.
Sign in.
Select the POST /brochure request and open its Cookies tab.
If there’s a yellow warning icon next to the session cookie, hover over it.
If the warning mentions SameSite, the fix is on the dev API side: set the session cookie to SameSite=None; Secure, or proxy the API through the dev server on 127.0.0.1 so the cookie becomes first-party.
It would also help to know:
Does the same sign-in work in Google Chrome on your machine?
What exact Cursor version is your colleague using? Check Cursor > About Cursor.
Let me know what DevTools shows, and we’ll take it from there.
Hey Renee. Thanks for the screenshot and the Request ID. That makes the picture clear.
The warning in DevTools on POST /brochure is the key part:
This Set-Cookie header didn’t specify a “SameSite” attribute and was defaulted to “SameSite=Lax”, and was blocked because it came from a cross-site response… The Set-Cookie had to have been set with “SameSite=None” to enable cross-site usage.
This means the session cookie from the API is coming back without a SameSite attribute, so the browser applies the default SameSite=Lax. Lax cookies aren’t sent on cross-site fetch requests, so when the page on 127.0.0.1 calls an API on another domain, it doesn’t get the cookie back. That’s why POST /user_role_allowed returns 401.
One important detail: you mentioned that the same login also fails in regular Google Chrome on your machine with “Your session has ended”. That confirms this isn’t specific to Cursor or the 3.23.23 update. It’s standard cross-site cookie behavior that Chromium applies the same way in both Chrome and the Cursor browser. Firefox works because it doesn’t enforce the default Lax behavior in the same way here.
Why it works for your colleague: they’re on Windows and a different build, so the environment is different. But since regular Chrome is also blocking the cookie on your side, the safer fix is on the API side instead of relying on a specific OS/version setup.
What should fix this everywhere:
Set the session cookie on the dev API with SameSite=None and Secure. None requires Secure, which means HTTPS. Your API is already on HTTPS, so that part should be fine.
Or proxy the API through the dev server on 127.0.0.1 so the cookie becomes first-party. Then the cross-site cookie rule won’t apply.
Either option should solve it. Let me know if anything still comes up after changing the cookie attribute, and we can dig further.