Urgent: Completely unable to use Cursor - ALL models broken, support unresponsive for 2 days

Support Help
1

I’m a paying and for the past 2 days I have been completely locked out of using ANY AI features in Cursor. Not a single model works. I’m dead in the water.

What’s happening:

  • ERROR_OPENAI – “Unable to reach the model provider”

  • Request ID: b17fdfc0-22c6-4612-90bb-a5da2f336f87

  • Started with external models (Claude, GPT, Gemini) not working

  • Now EVEN COMPOSER doesn’t work

  • NOTHING works. Zero. Nada.

What I’ve tried (LITERALLY EVERYTHING):

  • :white_check_mark: Ran network diagnostics - all clear

  • :white_check_mark: Switched between ALL models - none work

  • :white_check_mark: Not using VPN

  • :white_check_mark: Not using proxy

  • :white_check_mark: Disabled HTTP/2

  • :white_check_mark: Restarted Cursor multiple times

  • :white_check_mark: Restarted my entire computer multiple times

  • :white_check_mark: Updated to the latest version

  • :white_check_mark: Uninstalled and reinstalled the application

  • :white_check_mark: Disabled firewall/antivirus

  • :white_check_mark: Tried different networks

I have done everything. Back and forth, back and forth. Uninstall, reinstall, restart, retry. Over and over again. NOTHING HAS WORKED.

The support situation:

I’ve been emailing support since December 29th. After going through all their troubleshooting steps (which made things WORSE - Composer stopped working after the update they recommended), I was told it would be “escalated to the technical team.”

That was 2 days ago. No response. No timeline. Nothing.

Why this is critical:

I have multiple projects to deliver. Clients are waiting. Deadlines are approaching. And I’m completely blocked because the software I pay for simply doesn’t work.

I cannot work. Period.

What I need:

  1. An actual fix - not more “try restarting” suggestions

  2. A timeline - when will this be resolved?

  3. Accountability - if your infrastructure is breaking paying customers’ workflows, that’s on you

Has anyone else experienced this? Is there anyone from Cursor actually monitoring these forums?

I shouldn’t have to beg on a public forum to get basic functionality that I’m paying for. This is unacceptable.

System info:

  • Cursor version: Latest (updated as instructed)

  • Request ID: b17fdfc0-22c6-4612-90bb-a5da2f336f87

1 Like
Connection Error occurs depending on prompt content (Windows 11)
2

Since its Christmas and close to New years, support is more than likely having most of them away.

As for all models broken, no, they’re not. I am using 2.2.44 and it has been working fine.

As for monitoring, yes there are 1 or 2 from cursor.

1 Like
3

I’m aswell in the 2.2.44 version, still nothing

4

Show me your rules file

5

I have no rules configured:

  • No Project Rules

  • No Project Commands

  • No User Commands

Everything is empty/default. The error “Unable to reach the model provider” is still showing in the bottom right corner.

This is not a configuration issue on my end.

6

Pick a different model, what happens. Out of curiosity what are you trying to achieve? Have you started a brand new chat and a completely different brand new project?

8

Hey, thanks for the detailed report, and sorry for the inconvenience.

Based on ERROR_OPENAI and the fact that ALL models are failing, this looks unusual. These errors are usually specific to a single model.

A few extra checks:

  1. Open Developer Tools (Help > Toggle Developer Tools), go to the Console tab, and try sending a request to any model. Please paste any red console errors here if they show up.

  2. Try testing each model separately in a new chat:

    • Claude 4.5 Sonnet (Cursor’s default model)
    • GPT-5.2
    • Gemini Pro 3

    Which one was the first to stop working on December 29?

  3. Cursor Settings > Privacy > Output Channel. Do you see any errors there?

  4. What OS and region are you on?

We’re aware of an issue with Gemini Pro 3 at high context (~100k+ tokens), and the team is working on a fix. But if even Claude fails from the start, this may be a different, network-level issue.

Send the info above and I’ll escalate this to the team with full context right away.

1 Like
11

Thanks for looking into this.

Here are all the console errors from Developer Tools for each model:

Claude Opus 4.5:

[composer] ToolFormer: error in toolWrappedStream undefined ConnectError: [unavailable] Error
    at szc.$endAiConnectTransportReportError (workbench.desktop.main.js:12375:456006)
[AiService] streamResponse ConnectError: [unavailable] Error
[transport] Stream response error ConnectError: [unavailable] Error
{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.2.44', generationUUID: '575b497f-b5be-4027-9f08-fe67f19a0380'}
[transport] Automatic bug report submitted for unexpected connection error
[composer] Error in AI response: {"error":"ERROR_OPENAI","details":{"title":"Unable to reach the model provider"}}
[composer] Failed to get complete AI response
[composer] Chat incomplete or aborted

Gemini 3 Pro:

[composer] ToolFormer: error in toolWrappedStream undefined ConnectError: [unavailable] Error
[AiService] streamResponse ConnectError: [unavailable] Error
[transport] Stream response error ConnectError: [unavailable] Error
{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.2.44', generationUUID: '048707c1-cf05-4923-8d2c-5bc8fb31ac0b'}
[transport] Automatic bug report submitted for unexpected connection error
[composer] Error in AI response: {"error":"ERROR_OPENAI","details":{"title":"Unable to reach the model provider"}}
[composer] Failed to get complete AI response
[composer] Chat incomplete or aborted

GPT 5.2:

[composer] ToolFormer: error in toolWrappedStream undefined ConnectError: [unavailable] Error
[AiService] streamResponse ConnectError: [unavailable] Error
[transport] Stream response error ConnectError: [unavailable] Error
{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.2.44', generationUUID: '1aa54b4b-40f3-468b-a68c-cca812ecc660'}
[transport] Automatic bug report submitted for unexpected connection error
[composer] Error in AI response: {"error":"ERROR_OPENAI","details":{"title":"Unable to reach the model provider"}}
[composer] Failed to get complete AI response
[composer] Chat incomplete or aborted

Composer:

[transport] Stream response error ConnectError: [unavailable] Error
{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.2.44', generationUUID: '1b0bffc3-9cb9-46a6-93fd-1cee83303466'}
[transport] Automatic bug report submitted for unexpected connection error
[composer] Error in AI response: {"error":"ERROR_OPENAI","details":{"title":"Unable to reach the model provider"}}
[composer] Failed to get complete AI response
[composer] Chat incomplete or aborted

Answers to your questions:

  1. Console errors: All pasted above - identical ConnectError: [unavailable] across ALL models

  2. First model to fail: ALL of them failed simultaneously on December 29th. It wasn’t gradual.

  3. Privacy > Output Channel: Same errors appearing there

  4. OS and Region:

    • macOS (Apple MacBook, arm64 architecture)

    • Region: Vietnam

As you can see, every single model throws the exact same ConnectError: [unavailable] error. This is clearly not a model-specific issue - something is blocking ALL connections to your model providers from my end, but I have no VPN, no proxy, firewall disabled, and network diagnostics show everything is fine.

Could Vietnam be geo-blocked or throttled? Or is there something on your infrastructure side affecting users in this region?

Please escalate this urgently - I’ve been completely unable to work for 2+ days now with multiple project deadlines approaching.

12

are you using your own OpenAI API key?

13

I’m not using any AI API

15

Thanks for the detailed report. Based on the logs showing the same ConnectError: [unavailable] for all models, this looks like a DNS routing issue and Cloudflare SSL certificates.

We recently found that some DNS servers, especially Google DNS 8.8.8.8, can route requests through a problematic Cloudflare endpoint with invalid SSL certificates. That would explain why Network Diagnostics shows “OK”, but all models fail.

Please try:

  1. Change DNS servers on macOS:

    • System Settings > Network > Wi‑Fi/Ethernet > Details > DNS
    • Try Quad9: 149.112.112.112 and 9.9.9.9
    • Or Cloudflare: 1.1.1.1 and 1.0.0.1
    • Or your ISP DNS (remove any custom DNS)

  2. After changing DNS, fully restart Cursor

  3. If it still doesn’t help, try:

    sudo pkill -9 trustd

    Then restart Cursor.

Also, please send:

  • Which DNS servers are currently configured on your machine? (System Settings > Network > DNS)
  • The output from Cursor Settings > Privacy > Output Channel (if there are extra errors)

Let me know if switching DNS helps.

19

Update after more testing:

What I tried:

  • Changed DNS to Cloudflare (1.1.1.1 and 1.0.0.1)

  • Ran sudo pkill -9 trustd

  • Restarted Cursor

  • Removed ALL MCPs completely

Result: Still getting the same error, even with all MCPs removed:

Request ID: cb942d0f-b9ac-491f-a69c-43e2f7ffc575
ERROR_OPENAI - Unable to reach the model provider
ConnectError: [unavailable] Error

So it’s not the MCPs.

To summarize:

  • I’ve been in Vietnam for over a month, Cursor worked perfectly until 3 days ago

  • Nothing changed on my end (same network, same laptop, same location)

  • The problem appeared suddenly on December 29th

  • It works intermittently (2 prompts then dies)

  • I’ve now tried literally everything suggested

At this point I’ve exhausted all client-side troubleshooting. This has to be something on your infrastructure - maybe a change you deployed around December 29th that affected routing to certain regions?

I need this escalated to your backend/infrastructure team. I’m on day 3 with a broken $100/month subscription and multiple project deadlines at risk.

Please provide a timeline for resolution or escalation confirmation.

20

Thanks for the detailed info. Based on the logs and the fact that the issue started suddenly on December 29, this looks like a network routing problem.

To escalate this, I need some extra DNS and network diagnostics. Please run the following in your terminal:

# DNS resolution
nslookup api2.cursor.sh
dig api2.cursor.sh

# Route trace
traceroute api2.cursor.sh

# Current external IP
curl -v ifconfig.me

Also, please try connecting via a mobile hotspot (if you have one). This will help confirm whether the issue is with your ISP or something more global.

Send me the results and I’ll pass them to the infra team with priority.

21

nourmatine@Nours-MacBook-Air ~ % # DNS resolution

nslookup api2.cursor.sh

dig api2.cursor.sh

# Route trace

traceroute api2.cursor.sh

# Current external IP

curl -v ifconfig.me

zsh: command not found: #

Server: 1.1.1.1

Address: 1.1.1.1#53

Non-authoritative answer:

api2.cursor.sh canonical name = api2geo.cursor.sh.

api2geo.cursor.sh canonical name = api2direct.cursor.sh.

Name: api2direct.cursor.sh

Address: 98.89.132.230

Name: api2direct.cursor.sh

Address: 44.194.137.253

Name: api2direct.cursor.sh

Address: 54.167.139.47

Name: api2direct.cursor.sh

Address: 52.1.194.173

Name: api2direct.cursor.sh

Address: 44.208.52.43

Name: api2direct.cursor.sh

Address: 34.205.230.207

Name: api2direct.cursor.sh

Address: 44.212.116.29

Name: api2direct.cursor.sh

Address: 54.156.154.21

; <<>> DiG 9.10.6 <<>> api2.cursor.sh

;; global options: +cmd

;; Got answer:

;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 50093

;; flags: qr rd ra; QUERY: 1, ANSWER: 10, AUTHORITY: 0, ADDITIONAL: 1

;; OPT PSEUDOSECTION:

; EDNS: version: 0, flags:; udp: 1232

;; QUESTION SECTION:

;api2.cursor.sh. IN A

;; ANSWER SECTION:

api2.cursor.sh. 3501 IN CNAME api2geo.cursor.sh.

api2geo.cursor.sh. 3501 IN CNAME api2direct.cursor.sh.

api2direct.cursor.sh. 13 IN A 44.221.238.24

api2direct.cursor.sh. 13 IN A 100.51.245.199

api2direct.cursor.sh. 13 IN A 100.50.21.129

api2direct.cursor.sh. 13 IN A 100.51.52.105

api2direct.cursor.sh. 13 IN A 34.233.46.77

api2direct.cursor.sh. 13 IN A 23.23.24.184

api2direct.cursor.sh. 13 IN A 54.205.165.162

api2direct.cursor.sh. 13 IN A 35.175.35.212

;; Query time: 58 msec

;; SERVER: 1.1.1.1#53(1.1.1.1)

;; WHEN: Wed Dec 31 01:59:23 +07 2025

;; MSG SIZE rcvd: 218

zsh: command not found: #

traceroute: Warning: api2.cursor.sh has multiple addresses; using 3.220.28.60

traceroute to api2direct.cursor.sh (3.220.28.60), 64 hops max, 40 byte packets

1 192.168.1.1 (192.168.1.1) 3.391 ms 3.261 ms 3.735 ms

2 113.22.0.43 (113.22.0.43) 7.502 ms 4.729 ms 3.600 ms

3 *

1 Like
23

Im was happily coding away until Cursor informed me of a update. So it installed 2.3.15 after that Cursor Agent was completely bricked. None of the agents would work, all checks are OK.

DO NOT INSTALL 2.3.15 !!!

If you did, download 2.2.44 from the download site and install over the existing install, no need to uninstall first.

DO NOT INSTALL ANY UPDATES and disable AUTO Updates via: How do I prevent automatic updating? - #8 by minemindmedia

It seems to work for me or saved me at least.

Greetz

1 Like
24

Strongly agree, it worked fine before 10 minutes, but I performed an update → 2.3.15, all AI functions became unavailable, whether it was an old agent or a new agent, there was only one notification Agent Execution Timed Out, I used the network test tool, it showed that chat, agent connection was normal, I am trying to restore by downgrading the version

25

Normal use was restored by downgrading to 2.2.44, if you have the same problem, you can go to Download · Cursor Download 2.2.44 (latest)

26

Please start your own Bug Report thread. Don’t hijack another thread please.

29

hello? Still having the same problem

30

@deanrie