Grok Bot 0.30.0 (Windows) — setup fails with "Can't reach your computer", all network checks pass

Where does the bug appear (feature/product)?

Grok Bot

Describe the Bug

First-time setup never completes. The app shows the startup spinner, then fails with “Grok Bot couldn’t finish setting up — Can’t reach your computer right now. Check your connection and try again.”

I have verified this is not a local network or DNS issue — full evidence in Additional Information below.

Settings → Reset Agent Computer is not reachable, because the computer has never been created.

Steps to Reproduce

  1. Install Grok Bot 0.30.0 on Windows and sign in.
  2. Let first-run setup create the agent computer.
  3. Setup hangs, then fails with “Can’t reach your computer right now. Check your connection and try again.”
  4. Uninstall, delete leftover app data in %APPDATA% and %LOCALAPPDATA%, reinstall clean.
  5. Same failure on every attempt.

Expected Behavior

Setup creates the agent computer and the main Grok Bot interface loads.

Operating System

Windows 10/11

Version Information

Grok Bot 0.30.0 (Windows)

Additional Information

Network verification — all passing.

  • nslookup us8.cursorvm.com via ISP resolver (Reliance, India) returns 3.139.242.47, 18.223.226.110, 16.58.116.176
  • The same lookup against 1.1.1.1 returns the identical set of IPs — no resolver discrepancy, so the DNS workaround from forum thread 169007 does not apply here
  • No AAAA records exist for that host, so this is not an IPv6 fallback hang
  • curl -I against it over HTTPS returns HTTP/1.1 404 from awselb/2.0 — the expected response, confirming the path to your load balancer is reachable from my machine

Also ruled out:

  • Only one Grok Bot install present on the machine (no stale or duplicate install)
  • Fresh install today, not an auto-update from an earlier build

Since DNS, IPv6, and TCP/TLS reachability all check out from my end, the agent computer for my account appears to be stuck or never provisioned. Could you please rebuild/reset the Agent Computer associated with my account? Happy to provide my account email or logs privately.

Location: India (Kolkata) — mentioning in case region or capacity is a factor.

Possibly the same incident as the 0.30.0 macOS setup hang (thread 169981) and “Grok Bot couldn’t finish setup” (thread 170010).

Does this stop you from using Cursor

No - Cursor works, but with this issue

Possibly the same incident as the 0.30.0 macOS setup hang (thread 169981) and “Grok Bot couldn’t finish setup” (thread 170010).

Hey, thanks for the detailed report, the diagnostics are really thorough.

First, the good news: your Grok Bot computer on our side has been created and it’s working. The failure happens on the last step, when your machine tries to reach it. So this isn’t a “computer never provisioned” case, it does exist.

One thing your checks likely missed: the app doesn’t connect to us8.cursorvm.com itself, it connects to a separate per-computer subdomain under cursorvm.com (a long unique hostname tied to your computer). The apex resolves for you, but the per-computer lookup, based on the symptom, is failing somewhere on your resolver path.

The first and main question: are you using a VPN? With a VPN enabled, the tunnel’s DNS resolver might not resolve per-computer subdomains under cursorvm.com, and the lookup times out. If you are using a VPN, try:

  1. Turn off the VPN completely and rerun setup. If it works, that’s the cause.
  2. If you need the VPN, set up split tunneling or add *.cursorvm.com to exceptions. Important: the exception must cover DNS resolution, not just routing (many VPNs force all DNS through their resolver).

If you’re not using a VPN, check wildcard subdomain resolving under the same zone:

nslookup test123.us8.cursorvm.com

If that request fails as non-existent or times out, while us8.cursorvm.com works, the issue is subdomain resolving. Then try:

  1. Set a public DNS resolver in Windows: Settings > Network and internet > your connection > DNS server assignment > Manual, IPv4, 1.1.1.1 preferred, 8.8.8.8 alternate. Then fully restart Grok Bot.
  2. If that doesn’t help, connect via a phone hotspot from a different carrier. If setup works there, the issue is with your normal network’s resolver.

Let me know if you’re using a VPN and what nslookup shows, and we can decide the next step.

Fixed — thanks, your wildcard subdomain theory was exactly right.

Not using a VPN. Plain Reliance broadband.

nslookup test123.us8.cursorvm.com        (Reliance resolver)
*** can't find test123.us8.cursorvm.com: Query refused

nslookup test123.us8.cursorvm.com 1.1.1.1
Addresses: 3.139.242.47, 18.223.226.110, 16.58.116.176

So the apex resolves on Reliance but any subdomain under it gets REFUSED — not NXDOMAIN, the resolver is actively declining to answer.

One wrinkle that cost me some time: setting IPv4 DNS to 1.1.1.1 / 8.8.8.8 was not enough. My router advertises a Reliance IPv6 resolver (2405:201:8006:dc10::c0a8:1d01) over RA, and Windows preferred it over the IPv4 setting, so lookups still went to Reliance and still got refused. Disabling IPv6 on the adapter (or setting IPv6 DNS to 2606:4700:4700::1111 explicitly) was what actually made it take effect. Setup completed immediately after.

For anyone else hitting this from India: test a wildcard subdomain, not just the apex, and verify the Server line in nslookup output actually shows your new resolver — an IPv6 resolver from RA will silently override IPv4 DNS settings.

Great detective work, and thanks for coming back with a full write-up. The IPv6 find is especially valuable.

You nailed what often gets missed. Even if you set IPv4 DNS to 1.1.1.1 or 8.8.8.8, an IPv6 resolver coming in via router RA can quietly override that, so requests still go to the ISP resolver. Checking the Server line in nslookup output is the golden rule here.

Quick notes for anyone who runs into this in the same region:

  1. Test a wildcard subdomain like nslookup test123.us8.cursorvm.com, not just the apex. If the apex resolves but the subdomain gets Query refused, it’s the resolver.
  2. Change DNS for both IPv4 and IPv6, or disable IPv6 on the adapter, then confirm nslookup actually shows the new resolver in Server.

I’ll keep the thread open as a reference for others. If you get stuck again with the setup, post here and we’ll dig in.

Environment

Windows
Grok Bot 0.30.0
SuperGrok subscription

Problem

SuperGrok authentication and account linking completed successfully.

Before creating my first Bot, the Grok Bot UI worked normally. I could see my SuperGrok account information and usage.

After I created/configured my first Bot, the app started setting up the Grok Bot computer. After that, the entire app became a white screen with only a black loading dot in the center.

The UI never finishes loading, so I cannot access Settings, Recover, Update Agent Computer, or Reset Agent Computer.

Network troubleshooting already completed

I am located in mainland China, so I need to use a VPN/proxy to access Grok services.

I tested both Wi-Fi and mobile hotspot and got the same result.

DNS resolution for the Cursor VM domain works correctly and returns AWS IP addresses.

TCP port 443 connectivity test:

TcpTestSucceeded : True

My HTTPS proxy is running on:

127.0.0.1:7890

Proxy CONNECT succeeds.

TLS connection succeeds.

A HEAD request to the us8 Cursor VM domain successfully reaches AWS and returns:

HTTP/1.1 404 Not Found

Server: awselb/2.0

So basic DNS resolution, TCP 443 connectivity, proxy tunneling, TLS, and AWS load balancer reachability are working.

Important detail

The Grok Bot interface worked normally before the first Bot / Agent Computer setup. The white-screen problem only started after the Agent Computer initialization began.

Could you please check whether the Agent Computer associated with my Grok/SuperGrok account is stuck during provisioning or is unreachable?

Since the UI does not load, I cannot perform Recover or Reset myself.

If necessary, please rebuild/reset the Agent Computer from the backend.

I can provide my Grok account email privately if needed.

Screenshot of the white screen with the black loading dot is attached.

Hey @PEI_NAN_JIANG

I’ve tweaked some knobs in the backend and you should be back in action. If not, please raise a new thread. I’ll close this one as @Mayank_Kamani’s issue was solved.