Iam not able to get codebase indexing to work in Windows 11 cursor user build installations of versions 3.12, 3.13 and 3.14. It works fine in v3.11
After upgrading my cursor installation from v3.11 to 3.14, I see codebase indexing is failing with below error when we add a folder with files to cursor. When i uninstall the higher versions and go back to v3.11, doing the same steps (opening the same test folder with files in it immediately kickstarts codebase indexing and indexing completes to 100%)
2026-08-05 08:33:17.420 [info] Creating Indexing Repo client:
2026-08-05 08:33:17.421 [info] Creating repo client with backend url:
2026-08-05 08:33:17.421 [error] NoWorkspaceUriError
2026-08-05 08:33:17.815 [error] Error getting repo info: Failed to execute git
2026-08-05 08:33:17.815 [error] Error: Failed to execute git
at rae. (c:\Users\my_name\AppData\Local\Programs\cursor\resources\app\extensions\cursor-retrieval\dist\main.js:2:2057549)
at Generator.next ()
at r (c:\Users\my_name\AppData\Local\Programs\cursor\resources\app\extensions\cursor-retrieval\dist\main.js:2:2055104)
at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
2026-08-05 08:33:18.115 [error] Error getting repo info: Failed to execute git
2026-08-05 08:33:18.212 [warning] Failed to get symbolic ref: Failed to execute git
2026-08-05 08:33:18.229 [warning] Failed to read .git/HEAD: A system error occurred (ENOENT: no such file or directory, open ‘C:\Users\my_name\AppData\Local\Programs\cursor.git\HEAD’)
2026-08-05 08:33:18.953 [warning] Failed to get branches from remote: Failed to execute git
2026-08-05 08:33:19.503 [info] Privacy mode changed to 2; re-evaluating indexing watcher.
2026-08-05 08:33:19.503 [error] NoWorkspaceUriError
2026-08-05 08:33:40.398 [info] Creating Indexing Repo client: https://repo42.cursor.sh
2026-08-05 08:33:40.398 [info] Creating repo client with backend url: https://repo42.cursor.sh
2026-08-05 08:33:40.402 [info] Making keys for a new index
2026-08-05 08:33:40.925 [info] Using the default key from the server for a new index
2026-08-05 08:33:40.925 [info] InternalRepoInfo constructor with preferredEmbeddingModel 0
2026-08-05 08:33:41.161 [warning] Failed to get symbolic ref: Failed to execute git
2026-08-05 08:33:41.266 [error] Error getting repo info: Failed to execute git
2026-08-05 08:33:41.266 [error] Error: Failed to execute git
at rae. (c:\Users\my_name\AppData\Local\Programs\cursor\resources\app\extensions\cursor-retrieval\dist\main.js:2:2057549)
at Generator.next ()
at r (c:\Users\my_name\AppData\Local\Programs\cursor\resources\app\extensions\cursor-retrieval\dist\main.js:2:2055104)
at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
2026-08-05 08:33:41.273 [error] Error getting repo info: Failed to execute git
2026-08-05 08:33:42.279 [warning] Failed to get symbolic ref: Failed to execute git
2026-08-05 08:33:46.459 [info] Privacy mode changed to 2; re-evaluating indexing watcher.
2026-08-05 08:33:46.461 [info] Not creating an indexing watcher: no queryable index to build (indexing off or kill switch engaged) and no training telemetry to produce.
2026-08-05 08:33:46.462 [info] Using existing keys for an index
2026-08-05 08:33:46.462 [info] InternalRepoInfo constructor with preferredEmbeddingModel 0
2026-08-05 08:33:46.464 [info] Not creating an indexing watcher: no queryable index to build (indexing off or kill switch engaged) and no training telemetry to produce.
2026-08-05 08:33:46.464 [info] Using existing keys for an index
2026-08-05 08:33:46.465 [info] InternalRepoInfo constructor with preferredEmbeddingModel 0
For AI issues: which model did you use?
Model name (e.g., Sonnet 4, Tab…)
For AI issues: add Request ID with privacy disabled
Request ID: f9a7046a-279b-47e5-ab48-6e8dc12daba1
For Background Agent issues, also post the ID: bc-…
Additional Information
Add any other context about the problem here.
Does this stop you from using Cursor?
Yes - Cursor is unusable
Sometimes - I can sometimes use Cursor
No - Cursor works, but with this issue
The more details you provide, the easier it is for us to reproduce and fix the issue. Thanks!
What you are seeing is expected on current Cursor builds. We changed how agents search the codebase. Models got very good at grep / indexed search, so the older dedicated semantic search path was no longer helping in a meaningful way. Agents now use our Instant Grep tool plus normal file reads.
For most workflows, this should feel the same or better. Search runs against the files on your machine, so Cursor no longer builds a semantic index of the repo. We are no longer computing embeddings of your code or storing them on our servers for search. On some older versions, the indexing UI can still show progress. That does not mean a separate semantic search tool is available to the agent.
More detail, including .cursorignore / .gitignore and how local Instant Grep works, is here:
Could you please point me to the release note/changelog indicating this feature is dropped in latest releases? I tried to search but couldn’t find it release noted anywhere. Much appreciated!
it is working only if Settings → “Indexing & Docs” indicates Indexing progress and ultimately shows 100% completed status . And View → Output → “Cursor Indexing % Retrieval” drop down option shows handshake kickstarting and indexing starting and completing correctly as expected.
Not in the Model’s chain of thought (COT) ouput as you have pointed out above.
Sorry, I probably misunderstood you, and the decision to disable the built-in index has already been discussed above. You can continue to use external MCPs:
Interesting clarification. It seems the important distinction is between repository search and repository understanding. Grep and file reads can be very effective for locating relevant code, but architectural reasoning, cross module dependencies and debugging still require connecting evidence across the codebase. That’s an interesting direction for AI coding tools.
We’ve been exploring this problem with Meeba AI Brain using repository connectivity, code graph reasoning and MCP, specifically around architecture and evidence based debugging rather than semantic search alone.
