Cursor Not Working: Fixes for the Most Common Errors

CursorErrors & BugsUpdated May 16, 2026
Quick Answer

When Cursor stops working, the most common causes are a failed AI model connection, an expired API key or subscription, a corrupted extension state, or a VS Code compatibility issue. Start by checking your Cursor account status at cursor.com, then reload the window with Ctrl+Shift+P → Reload Window. Most Cursor issues are resolved by signing out and back in or reinstalling the app.

Why Cursor Stops Working

Cursor combines a code editor with AI capabilities, so issues can come from either layer. Understanding which layer is failing saves diagnostic time.

  • AI connection failures — Problems reaching Cursor's AI backend: expired subscription, network blocks, or backend outage
  • Editor issues — Problems in the VS Code foundation: extension conflicts, corrupted settings, or cache issues
  • Authentication issues — Expired session, wrong account, or subscription not synced
  • Performance issues — Large codebases, low memory, or indexing failures

Step-by-Step Fix

1. Check Cursor's Status and Your Subscription

Go to cursor.com/account and confirm your subscription is active. Also check if Cursor has a status page or posts updates on their Twitter/X account for backend outages.

If your free trial or subscription has expired, AI features will stop working. This is the most common cause of sudden failures in Cursor.

2. Reload the Window

The fastest general fix for most Cursor issues:

  1. Press Ctrl+Shift+P (Cmd+Shift+P on Mac)
  2. Type Reload Window
  3. Press Enter

This resets the extension host and AI connection without closing your editor.

3. Sign Out and Sign Back In

A stale authentication token causes many persistent Cursor errors:

  1. Open command palette: Ctrl+Shift+P
  2. Type Cursor: Sign Out and run it
  3. Wait for the sign-out to complete
  4. Sign back in with the same account used for your subscription
  5. Check if AI features work after signing in

4. Check Your Network and VPN

Cursor's AI features require outbound HTTPS connections to Cursor's API servers:

  • Disable any VPN and retry — VPN endpoints are frequently blocked or throttled for AI services
  • Test on a different network (phone hotspot) to rule out network-level blocks
  • If on a corporate network, AI traffic may be filtered — check with your IT team

5. Update Cursor

Open Help → Check for Updates and install any available updates. Cursor releases frequent bug-fix versions, and running an outdated version is a common cause of compatibility issues.

6. Clear Cursor's Cache

If reloading and signing in do not help, clear the cache:

Mac/Linux:

~/.cursor/cache
~/.cursor/logs

Windows:

%APPDATA%\Cursor\Cache
%APPDATA%\Cursor\logs

Close Cursor, delete these folders, and reopen.

7. Disable Conflicting Extensions

Extensions can conflict with Cursor's AI features. Disable all non-essential extensions through the Extensions panel and test. Re-enable them one at a time to find any conflicts.

8. Reinstall Cursor

If nothing else works, a clean reinstall resolves corrupted installations:

  1. Uninstall Cursor
  2. Delete any remaining Cursor folders from your user directory
  3. Download the latest version from cursor.com
  4. Install and sign in fresh

Common Mistakes to Avoid

  • Not checking subscription status first — If your trial ended, no amount of troubleshooting will restore AI features. Always check this first.
  • Assuming a VS Code extension will work identically in Cursor — Most VS Code extensions work, but some conflict with Cursor's AI layer. Extensions that modify the editor's language server are the most likely to cause problems.
  • Ignoring the .cursorignore file — Including massive generated files or node_modules in AI context slows Cursor down dramatically and can cause timeouts that look like crashes.
  • Running Cursor and VS Code simultaneously on large projects — Both applications indexing the same codebase simultaneously can exhaust memory on lower-spec machines.

Cursor on Windows vs Mac vs Linux

A few platform-specific issues are worth knowing:

Windows: Antivirus software sometimes blocks Cursor's AI network requests. Add Cursor to your antivirus allow list if you see frequent connection failures.

Mac: After macOS updates, Cursor occasionally loses keychain access for stored credentials. If you are prompted to sign in repeatedly, go to Keychain Access, search for Cursor, and delete any stored items, then sign in fresh.

Linux: Cursor on Linux sometimes has issues with the system keyring for credential storage. If login keeps failing, try running Cursor with --password-store=basic flag to use a simpler credential storage method.

Why This Happens

Cursor's AI features depend on an uninterrupted chain: valid subscription → authenticated session → outbound HTTPS access to Cursor's API servers → model availability on Cursor's backend. A failure at any point in this chain looks like "Cursor stopped working." Because Cursor is built on Electron and VS Code, issues can also originate in the editor layer — corrupted extension state, conflicting extensions, or a VS Code engine incompatibility — which is unrelated to your subscription or network. The reload window command (Ctrl+Shift+P → Reload Window) clears the editor layer without closing your files, making it the fastest first test.

FAQ

Q: How many AI requests does Cursor allow per month on the free trial?

Cursor's free trial includes approximately 50 slow requests (using less capable models) and a limited number of fast requests (using GPT-4 or Claude). When the trial ends or limits are hit, AI features stop working silently — autocomplete disappears and the chat returns connection errors. Check cursor.com/account to see your exact usage and trial status. After the trial, a Pro plan at $20 per month restores fast requests.

Q: Cursor says "context length exceeded." How do I fix it?

Context length errors occur when the amount of code Cursor has included in the AI request exceeds the model's token limit. Fix by reducing the @codebase references in your prompt, excluding large generated files via .cursorignore, breaking your request into smaller scoped questions, or switching to a model with a longer context window in Cursor settings. Files like package-lock.json and yarn.lock should always be excluded from AI context.

Q: Why does Cursor's chat work but autocomplete does not, or vice versa?

Chat and autocomplete are separate AI features that can fail independently. Autocomplete uses a different model endpoint and a different token budget than chat. If autocomplete is off but chat works, check Settings → Editor → AI Autocomplete to confirm it is enabled and that your file type is not excluded. If chat is off but autocomplete works, the issue is likely the chat model's endpoint being temporarily unavailable — reload the window to clear this state.

Related Articles

Additional FAQ

Q: How do I know if the problem is on my end or the platform's side? Check the platform's official status page first — most services maintain a public status page that shows current incidents and outages. If no incident is posted and the problem only affects your account (not reported widely on Reddit or Twitter), it is likely a local issue. Testing in incognito mode and on a different network also helps isolate whether the problem is browser-specific, network-specific, or account-specific.

Related Articles

View all Cursor guides

Cursor · Errors & Bugs

More Cursor errors & bugs guides

Browse all guides in this category to troubleshoot related issues faster.

Browse all guides →

Frequently Asked Questions

A connection error to the AI model in Cursor usually means one of three things: your Cursor subscription has expired or your free trial has ended, your internet connection is blocking requests to Cursor's API endpoints, or there is a temporary outage on Cursor's backend. Check your account status at cursor.com/account. If your subscription is active, try disabling your VPN, switching networks, and restarting Cursor. If you are on a corporate network, the AI requests may be blocked by a firewall — test on a personal network or hotspot.

Related Guides

Continue with nearby guides in the same topic to rule out adjacent causes faster.

ChatGPT error 403 – How to Fix

A ChatGPT 403 error means access was forbidden — the most common causes are expired session tokens, browser extensions blocking requests, or VPN/proxy IP mismatches. Fix it by refreshing the page, clearing cookies for chat.openai.com, and retrying in incognito mode; if the error appears on long prompts, split them into smaller parts under 2,000 words.

ChatGPT Error in Message Stream: How to Fix It

The 'Error in message stream' in ChatGPT means the real-time streaming connection that delivers words to your screen was interrupted mid-response — not that your message was lost. This error appears in roughly 60–70% of cases due to an unstable network connection or a VPN endpoint throttling the persistent connection; switching to a hotspot and disabling VPN resolves it for most users.

How to report a ChatGPT bug effectively (what to include)?

To report a ChatGPT bug effectively, include the exact error text, a timestamp, your OS and browser/app version, and clear steps to reproduce — reports missing even one of these 4 elements are resolved 2–3x slower by support. Start with a clean session (sign out, clear cache/cookies, disable extensions) to confirm the bug is reproducible, then submit via help.openai.com.