Cursor is an AI-first code editor. Developers use it to write, refactor, and debug code with an assistant that understands the whole project, autocompletes across files, and answers questions in a chat panel alongside the editor. Because that intelligence runs in the cloud, Cursor depends on more than your laptop: it needs its own backend services and the upstream model providers it routes requests to. When any link in that chain stalls, the editor still opens but the AI features go quiet, which is the moment most people start searching whether Cursor is down. The live verdict and recent user reports at the top of this page tell you, right now, whether the problem is on Cursor’s side or yours.
Common reasons Cursor stops working
- Backend or model-provider outage. Cursor relies on upstream AI providers. If one degrades, completions and chat slow down or fail even though the app itself loads fine.
- Sign-in and license sync. Expired sessions, auth token problems, or subscription checks not syncing can lock you out of Pro features or the assistant entirely.
- Rate limits and quota. Heavy use, or being on a busy plan tier during peak hours, can throttle requests so responses queue or time out.
- Indexing and context failures. Large repositories that fail to index leave the model without project context, so answers get vague or autocomplete stops firing.
- Network and firewall blocks. Corporate proxies, VPNs, or local firewalls often block the API endpoints Cursor needs while normal browsing still works.
- Stale install or extension conflicts. An outdated build or a misbehaving extension can break features that the servers are serving correctly.
How to fix Cursor when it is not working
- Check the verdict above first. If the status at the top of this page shows a confirmed outage, the issue is on Cursor’s end and the fix is to wait. Skip the local troubleshooting.
- Reload the window. Use the command palette to reload, or fully quit and reopen Cursor. This clears most transient session and indexing glitches.
- Re-authenticate. Sign out and back in to refresh your token and re-sync your subscription, then confirm the assistant responds again.
- Test your network. Disable a VPN or proxy, switch to a different connection, and confirm you are not behind a firewall that blocks Cursor’s API traffic.
- Update and check extensions. Install the latest version and disable recently added extensions to rule out conflicts.
- Re-index your project. Trigger a fresh codebase index so the model regains full context, especially after switching branches or large changes.
- Still stuck? Work through our checklist for what to do when an AI tool is down while you wait for service to recover.
Is Cursor down, or is it just you?
The quickest way to tell them apart is the signal at the top of this page combined with the user reports. If many people are reporting failures at the same time, the outage is almost certainly Cursor’s, not yours. If reports are quiet but you cannot get a response, the cause is more likely local: your account, your network, or your install. A simple tell is whether the editor loads but only the AI features fail. That pattern usually points to a backend or model-provider problem rather than something on your machine, and the steps above will not change much until service returns.
How we check Cursor
We combine two independent signals so you are not relying on a single source. First, we read Cursor’s official status page to capture what the company itself is reporting about its services. Second, we run our own reachability check from outside Cursor’s network, hitting the service directly to confirm whether it actually responds for a real user. When those two signals agree, the verdict is straightforward. When they disagree, that gap is useful information in itself, and we surface it rather than hide it. Full details of how we collect, weight, and time-stamp these checks are on our methodology page.
Alternatives to Cursor during an outage
If Cursor is down and you need to keep shipping, a few peer tools cover similar ground. GitHub Copilot brings AI completion and chat into most major editors. Windsurf is another AI-native editor with project-wide context. Copilot can help with explanations and quick code questions in a pinch. Before you switch, check that backup’s own status page first, since AI tools often share upstream model providers and can be affected by the same incident. Confirm it is healthy on its own before you move your work over.
How Cursor outages actually show up
Cursor is a desktop code editor, so the part you type in keeps working even when its servers do not. Plain file editing, search, and your local terminal run offline, but anything that calls a model, including the Tab autocomplete, the Composer, and the agent, depends on Cursor’s backend being reachable. When that link drops you typically see a “Connection failed” banner, requests that spin and then time out, or chat replies that never arrive.
Two distinct things can break independently. The model backend can be degraded while sign-in and the dashboard are fine, or your account can fail to authenticate while the models are healthy. A third common pattern is codebase indexing stalling, which leaves the agent without context even though chat technically responds.
- Tab and Composer hang or return “Connection failed”
- Sign-in loops or a session that will not refresh
- Indexing stuck, so answers ignore your repository
- Request usage limits reached, blocking further model calls