Updated June 2026
When Cursor can’t index your project, its AI loses context and answers get worse. Indexing usually stalls on large repos, ignored paths or a corrupt cache.
⚡ Quick fix
- Start with check the indexing status.
- Start with exclude heavy folders.
- Start with re-index from scratch.
- Start with check your connection and sign-in.
What this problem means
When Cursor can’t index your project, its AI loses context and answers get worse. Indexing usually stalls on large repos, ignored paths or a corrupt cache.
1. Check the indexing status
Open the Cursor settings panel and look at the codebase indexing indicator. If it’s stuck at a percentage, it’s likely a large-file or memory issue.
2. Exclude heavy folders
- Add
node_modules,dist,buildand large binaries to.cursorignore. - This dramatically speeds up embedding and avoids timeouts.
3. Re-index from scratch
In Settings > Features > Codebase Indexing, delete the index and rebuild it. A fresh index fixes most corruption.
4. Check your connection and sign-in
Embeddings are computed server-side. A dropped connection or expired session pauses indexing — sign out and back in.
5. Reduce repo size
Extremely large monorepos can exceed limits. Open a focused subfolder as the workspace instead of the entire repo.
Diagnostic checklist before you escalate
Agent and coding-assistant failures span model access, repository context, permissions, tool execution, terminal state, and usage limits. Start with a bounded task and a clean workspace. Review every proposed command and diff, especially when the agent can modify files or call external services.
- Confirm the selected model and plan support agent or tool use.
- Open the correct project and refresh its index or repository context.
- Check pending permission prompts, terminal errors, and ignored files.
- Retry with a small task that names the file, desired behavior, and acceptance check.
- Review diffs and tests before accepting changes or allowing destructive commands.
| Test | What the result tells you | Next move |
|---|---|---|
| Official status page reports an incident | The service is affected beyond your device | Pause local resets and monitor recovery |
| Private window works | Normal browser data or an extension is involved | Clear site data and enable extensions one by one |
| Another network works | DNS, VPN, proxy, firewall, or filtering is involved | Review the original network configuration |
| Failure follows the account everywhere | Account, plan, quota, or service-side state is likely | Collect evidence and contact official support |
Verify the agent with a bounded, reversible task
Test Cursor AI Not Indexing Codebase: How to on a small task that has an obvious expected result, such as changing one label, explaining one function, or adding a focused validation check. Give the agent the relevant file and acceptance condition. A healthy run should read the right context, request necessary permission, make only the intended change, and report how it verified the result.
Inspect the complete diff before accepting it. Then run the repository’s formatter, type checker, and focused tests yourself. If the agent claims success without a diff or test evidence, treat the task as incomplete. Only after this bounded test should you allow broader edits, terminal commands, package changes, or access to external services.
- The agent uses the intended repository and files.
- Permission prompts appear before consequential actions.
- The diff is limited to the requested behavior.
- Tests and type checks pass independently.
- Reverting the test change is straightforward.
Keep a short note of the working configuration and the date of the test. Products, models, browser versions, limits, and safety policies change over time, so a previously successful workaround may later become obsolete. Prefer current official documentation over old forum instructions, and reverse temporary diagnostic changes once testing is complete. This gives you a reliable baseline without leaving extensions disabled, security controls weakened, or experimental settings enabled indefinitely. Recheck the baseline after major updates before assuming an older failure has returned for the same reason. When possible, save a screenshot or sanitized log from the successful test so you can compare future behavior without relying on memory alone during later troubleshooting.
Also confirm that the workaround does not create a second problem. Check saved projects, account history, notifications, downloads, and connected integrations after the test. A change that restores one button while breaking synchronization or access elsewhere is not a complete fix. If several people use the same account, workspace, or network, ask one other person to repeat the safe test before applying the change broadly. Document that final result alongside the original symptoms for future reference.
When none of the fixes work
Repeat the smallest failing action once and record the exact local time and time zone. Note the product, model or feature, account plan, browser or app version, operating system, and whether the same action works in a private window, on another device, or on another network. This evidence is much more useful than saying the tool is “still broken.”
Use the provider’s official support channel. Include a screenshot with sensitive information removed and list the steps already tested. For developer tools, add sanitized request and response details, correlation IDs, and SDK versions. Never send passwords, one-time codes, API keys, session cookies, private repository contents, or complete payment information.
FAQ
Why is indexing so slow? Large or binary files. Use .cursorignore to skip them.
Does indexing send my code to the cloud? Embeddings are processed remotely; check Cursor’s privacy mode if that’s a concern.
Bottom line: ignore heavy folders and re-index — that fixes the vast majority of cases.
Bottom line: Work from the least disruptive test to the most specific one. Confirm service health, isolate session and network variables, then escalate with clean evidence instead of repeating the same failing action.